Download platform/modules/amazon_profit.py from fsanyoto/loopable: direct link, hf CLI and curl.
- Browser
- Download file 196 kB
-
https://huggingface.co/spaces/fsanyoto/loopable/resolve/9291c0100f4e7b8da2efcbbf1ebdea72ee1663e7/platform/modules/amazon_profit.py
- Command line
-
hf download hf://spaces/fsanyoto/loopable@9291c0100f4e7b8da2efcbbf1ebdea72ee1663e7/platform/modules/amazon_profit.py
-
curl -L -o amazon_profit.py https://huggingface.co/spaces/fsanyoto/loopable/resolve/9291c0100f4e7b8da2efcbbf1ebdea72ee1663e7/platform/modules/amazon_profit.py
196 kB
"""Command Center: one row per Amazon SKU with Odoo cost, profit, ads, stock, returns, health and score.
Net proceeds = net sales minus fees minus sponsored ads, from the `amzn_economics` mirror.
COGS follows incoming receipt valuations and posted landed costs through FIFO.
The module owns its table (`ut_amazon_sku_profit`). It is not a connector table, so people
edit the owned columns (Odoo product, units per SKU, owner, status, due date). A refresh
rewrites only the computed columns, keyed by SKU, so owned cells belong to the SKU and
survive every period switch.
`ut_amazon_asin_profit` and `ut_amazon_parent_profit` roll the SKU rows up on every refresh.
Projects retain their own verified period totals and Impact in the tenant projects store.
`validate()` ties each net proceeds component to independent Amazon amounts within 2% over settled
days, with order charges on purchase dates and monthly storage in its service month.
A refresh runs it for the last complete month and writes no figures when it
fails. A passing refresh stores the verified range and the day the rest is due to settle.
Owned items (R10, R31): a refresh opens one item on a Losing or Thin SKU with no status, for the
default owner, due in 14 days. It reopens a failed fix 30 days after the status was set. The
`status_set` column records the status and the date a refresh first saw it, because the
activity log is delivered async and pruned, so a refresh cannot read a status date from it.
"""
from __future__ import annotations
import atexit
import datetime as dt
import hashlib
import json
import math
import re
import shutil
import tempfile
import threading
from decimal import Decimal, ROUND_HALF_UP
from pathlib import Path
from modules import amazon_sku_metrics as M
from modules import amazon_part_dates as _part_dates
import core.user_tables as _registry
from core import module_tables
from core import amazon_rivals as _rivals
TABLE_KEY = "ut_amazon_sku_profit"
TABLE_LABEL = "Project management"
PROJECTS_KEY = "ut_amazon_projects"
PROJECTS_LABEL = "Projects"
ASIN_KEY = "ut_amazon_asin_profit"
ASIN_LABEL = "Child ASIN"
PARENT_KEY = "ut_amazon_parent_profit"
PROFIT_MAX_FIELDS = 100 # A module table declares its field cap; a connector table stays exempt.
PARENT_LABEL = "Parent ASIN"
# The four tables are the module's, not a person's: they take no MAX_TABLES slot (D-3284 spirit).
_registry.register_module_tables(TABLE_KEY, PROJECTS_KEY, ASIN_KEY, PARENT_KEY)
# S82-T70 (R33): refresh adds the SKU, ASIN and Parent ASIN rows; a person edits them but never adds
# one. Never `register_locked_records`: that turns `records_mutable` False and the edit doors then
# refuse Owner, Status, Due date and Odoo product. Projects stays open.
NO_ADD_KEYS = (TABLE_KEY, ASIN_KEY, PARENT_KEY)
NO_ADD_LINE = "Rows here come from Amazon. Edit a row instead."
module_tables.register_no_add_records(*NO_ADD_KEYS, line=NO_ADD_LINE)
SETTINGS_KEY = "amazon_profit_settings"
COST_HISTORY_KEY = "amazon_profit_cost_history"
VIA = "amazon_profit"
WRITER = "automation"
TENANT_ZERO = "royal-imports"
PERIOD_DEFAULT = "90d"
PERIOD_DAYS = {"30d": 30, "90d": 90}
THIN_DEFAULT = 0.10
STATUSES = ("Open", "Repriced", "Ads cut", "Delisted", "Keep", "COGS wrong")
#: A fix that reopens when the SKU still loses money or runs thin 30 days on. Keep never reopens.
FIX_STATUSES = ("Repriced", "Ads cut", "Delisted", "COGS wrong")
BANDS = ("Losing", "Thin")
NEEDS_REVIEW, LINKED, WRONG_MATCH = "Needs review", "Linked", "Wrong match"
#: The old names stay for callers. A decision word is a Link status option.
CONFIRMED, REJECTED = LINKED, WRONG_MATCH
MATCH_STATES = (NEEDS_REVIEW, LINKED, WRONG_MATCH)
#: Option words a stored table may still hold. `ensure_table` rewrites them, cells included.
LEGACY_MATCH_STATES = {"Confirmed": LINKED, "Rejected": WRONG_MATCH}
LINK_PURPOSE = "the Amazon SKU to its Odoo product for cost and profit"
DUE_DAYS = 14
REOPEN_DAYS = 30
TEMPLATE_KEY = "amazon_sku_profit"
ASIN_TEMPLATE_KEY = "amazon_asin_profit"
PARENT_TEMPLATE_KEY = "amazon_parent_profit"
PROJECT_STATUSES = ("Open", "Done")
PROJECT_HEALTH = ("On track", "Due this week", "Overdue", "Done")
DUE_SOON_DAYS = 7
PICKER_MONTHS = 12
SETTLEMENT_TOLERANCE = 0.02
#: Amazon releases funds about 7 days after delivery, so a month settles about two weeks after it ends.
SETTLEMENT_DEFERRAL_DAYS = 14
ECONOMICS = "amzn_economics"
SETTLEMENTS = "amzn_settlements"
LISTINGS = "amzn_listings"
LOAN_REPAYMENT = "Automated Loan Repayment"
WITHHELD_TAX = "ItemWithheldTax"
_ECONOMICS_REQUIRED = ("msku", "start_date", "sales_net_product_sales_amount",
"sales_net_units_sold", "fees_total_amount")
_SETTLEMENT_REQUIRED = ("amount", "posted_date", "transaction_type", "amount_type")
_ADS_TOTAL = re.compile(r"ads_[a-z0-9_]+_amount")
_MONTH = re.compile(r"(\d{4})-(\d{2})")
_LOCK = threading.Lock()
class SourceUnavailable(RuntimeError):
"""The Amazon mirror lacks a table or column this module reads."""
class TableRefused(RuntimeError):
"""The store refused the module table; the message names the cause and the fix."""
def _field(key, label, ftype, description, *, computed, **extra):
out = {"key": key, "label": label, "type": ftype, "origin": "loopable",
"valueAuthority": "computed" if computed else "user-entered",
"scope": "tenant", "description": description}
if ftype == "currency":
out["format"] = {"decimals": 2}
out.update(extra)
return out
def _progress_fields(description, *, key="projects", label="Projects", empty_label="No projects"):
"""Project progress and its completed-count label, written by the Projects doors."""
return (_field(key, label, "pct", description, computed=True,
format={"display": "progress", "labelKey": key + "_label", "emptyLabel": empty_label}),
_field(key + "_label", label + (" completed" if key == "projects" else " label"), "text",
"Completed projects of the counted ones." if key == "projects" else "Open PDP workflow stage and progress.",
computed=True))
HISTORY_FIELD = _field("history", "History", "text", "Open the row's recorded changes.", computed=True)
ASIN_APP_KEYS = ("asin", "projects", "pdp_update", "trend_score", "search_trend", "market_share_trend",
"click_share_trend", "exposure", "funnel_score", "pdp_last_updated", "title_last_updated",
"bullets_last_updated", "images_last_updated", "aplus_last_updated", "history")
SEARCH_TRENDS = (("search", "search_trend", "Search trend", "int"),
("market_share", "market_share_trend", "Market share trend", "pct"),
("click_share", "click_share_trend", "Click share trend", "pct"))
EXPOSURES = {"under": "Under-exposed", "match": "Matched", "over": "Over-exposed",
"few": "Few", "none": "Not seen"}
PDP_PART_DATE_KEYS = tuple(part + "_last_updated" for part in ("title", "bullets", "images", "aplus"))
PDP_CELL_KEYS = ("pdp_update", "pdp_update_label", "pdp_update_more") + PDP_PART_DATE_KEYS
_PDP_PROGRESS_FIELDS = _progress_fields("Progress of the earliest open PDP workflow.", key="pdp_update",
label="PDP update", empty_label="No open workflow")
_SIGNAL_FIELDS = (
_field("trend_score", "Trend score", "int", "Search trend score from 0 to 100.", computed=True),
*(_field(key, label, "pct", "Change against the same reported months last year.", computed=True)
for _, key, label, _ in SEARCH_TRENDS),
_field("exposure", "Exposure", "select", "Sales or click share against impression share.",
computed=True, options=list(EXPOSURES.values())),
_field("funnel_score", "Funnel score", "int", "Search-weighted funnel score, with 5 at the market.",
computed=True, format={"decimals": 1}),
)
_ASIN_DATE_FIELDS = tuple(_field(key, label, "date", "Latest recorded listing or part change.", computed=True)
for key, label in zip(("pdp_last_updated",) + PDP_PART_DATE_KEYS,
("PDP last updated", "Title last updated", "Bullets last updated",
"Images last updated", "A+ last updated")))
_SIGNAL_COMPANIONS = tuple(field for kind, key, label, now_type in SEARCH_TRENDS for field in (
_field(key + "_series", label + " series", "text", "Every reported historical month, with raw values, ASIN gaps and reasons.", computed=True),
_field(key + "_now", label + " now", now_type, "Searches in the newest complete store month." if kind == "search"
else "Share aggregated over the twelve-month calculation window.", computed=True),
_field(key + "_reason", label + " reason", "text", "Why the year comparison is unavailable.", computed=True),
_field(key + "_now_reason", label + " now reason", "text", "Why the current value is unavailable.", computed=True),
)) + tuple(_field(key + "_reason", label + " reason", "text", "Why this signal is unavailable.", computed=True)
for key, label in (("trend_score", "Trend score"), ("funnel_score", "Funnel score"), ("exposure", "Exposure")))
_ASIN_COMPANIONS = (_PDP_PROGRESS_FIELDS[1:]
+ (_field("pdp_update_more", "More PDP workflows", "int", "Other open PDP workflows.", computed=True),)
+ _SIGNAL_COMPANIONS
+ (_field("exposure_label", "Exposure detail", "text", "Exposure multiple and comparison base.", computed=True),
_field("pdp_last_updated_reason", "PDP date reason", "text", "Why Amazon's listing date is unavailable.", computed=True)))
_ASIN_SOURCE_KEYS = tuple(f["key"] for f in _SIGNAL_FIELDS + _SIGNAL_COMPANIONS) + (
"exposure_label", "pdp_last_updated", "pdp_last_updated_reason")
def _metric_fields(skip=()):
"""The sibling metric module's columns, as computed fields of this module."""
return tuple(_field(key, label, ftype, description, computed=True, **dict(extra or {}))
for key, label, ftype, description, extra in M.METRIC_COLUMNS if key not in skip)
#: The profit columns: figures from the economics mirror and Odoo cost, written by `cells`.
_ECON_FIELDS = (
_field("sku", "SKU", "text", "Amazon seller SKU.", computed=True),
_field("asin", "ASIN", "text", "Amazon product id.", computed=True),
_field("title", "Title", "text", "Listing title.", computed=True),
_field("period", "Period", "text", "Dates these figures cover.", computed=True),
_field("units_sold", "Units sold", "int", "Net units sold in the period.", computed=True),
_field("net_sales", "Net sales", "currency", "Sales after refunds.", computed=True),
_field("fees", "Fees", "currency", "Amazon fees.", computed=True),
_field("sponsored_ads", "Sponsored ads", "currency", "Ad spend.", computed=True),
_field("net_proceeds", "Net proceeds", "currency", "Net sales minus fees and ads.",
computed=True),
_field("sales_note", "Sales note", "text", "Why sales, units, fees or proceeds are blank.", computed=True),
_field("cogs_per_unit", "COGS per unit", "currency", "Receipt purchase and landed cost per Amazon unit.",
computed=True),
_field("cogs_total", "COGS total", "currency", "COGS per unit times units sold.",
computed=True),
_field("net_profit", "Net profit", "currency", "Net proceeds minus COGS.", computed=True),
_field("margin", "Margin", "pct", "Net profit over net sales.", computed=True),
_field("cost_note", "Cost note", "text", "Why COGS, profit or margin is blank.", computed=True),
_field("band", "Band", "select", "Losing or thin.", computed=True, options=list(BANDS)),
_field("cost_receipt", "Receipt", "text", "Receipts supplying these units in FIFO order.", computed=True),
_field("cost_source", "Cost source", "text", "Landed cost, purchase price or missing coverage.", computed=True),
_field("cost_coverage", "Cost coverage", "text", "Covered Odoo units and allocation method.", computed=True),
_field("cost_history", "Cost history", "text", "Saved receipt observations, appended only on refresh.",
computed=True),
_field("odoo_channel_sales", "Odoo channel sales", "currency", "Confirmed Fisch and Royal sales excluding Amazon.",
computed=True),
_field("odoo_channel_profit", "Odoo channel profit", "currency", "Odoo margin on those sales in this period.",
computed=True),
_field("odoo_channel_margin", "Odoo channel margin", "pct", "Odoo margin divided by those sales.",
computed=True),
_field("odoo_margin_source", "Odoo margin source", "text", "Scope or missing Odoo margin data.", computed=True),
)
_ECON_KEYS = tuple(f["key"] for f in _ECON_FIELDS)
#: The row's main listing picture: an identity cell, set after the rollup and never summed.
PICTURE_FIELD = _field("picture", "Picture", "image", "Main Amazon listing picture.", computed=True)
#: Why a health label, the score, the priority or a trend is blank. It sits after the trends.
HEALTH_NOTE_FIELD = _field("health_note", "Health note", "text", "Why health, score or a trend is blank.",
computed=True)
#: Computed columns: a refresh owns them and the pre-set lock keeps people out of them.
COMPETITOR_FIELDS = tuple(_field(k, label, t, d, computed=True, **dict(x))
for k, label, t, d, x in _rivals.ROW_COLUMNS)
PARENT_COMPETITOR_FIELDS = tuple(_field(k, label, t, d, computed=True, **dict(x))
for k, label, t, d, x in _rivals.PARENT_COLUMNS)
COMPETITOR_KEYS = _rivals.ROW_KEYS + _rivals.PARENT_KEYS
_SKU_PROJECT_FIELDS = _progress_fields("Average progress of the projects on this SKU, its ASIN and its parent ASIN.")
COMPUTED_FIELDS = (_ECON_FIELDS[:1] + _SKU_PROJECT_FIELDS[:1] + (HISTORY_FIELD, PICTURE_FIELD) + _ECON_FIELDS[1:] + _metric_fields()
+ (HEALTH_NOTE_FIELD,) + _SKU_PROJECT_FIELDS[1:]
+ COMPETITOR_FIELDS)
#: Owned columns: people edit them and a refresh never writes them.
OWNED_FIELDS = (
_field("odoo_product", "Odoo product", "text", f"Links {LINK_PURPOSE}.", computed=False),
_field("units_per_sku", "Units per SKU", "int", "Odoo units in one Amazon unit.",
computed=False, format={"decimals": 4}),
_field("owner", "Owner", "user", "Who acts on this SKU.", computed=False),
_field("status", "Status", "select", "Where the action stands.", computed=False,
options=list(STATUSES)),
_field("due", "Due date", "date", "When the action is due.", computed=False),
_field("match_state", "Link status", "select", f"Linked links {LINK_PURPOSE}.",
computed=False, options=list(MATCH_STATES)),
# S82-T71: Data Entry writes these back here. Restock reads carton CBM over units per carton.
_field("carton_size", "Carton size", "text", "Carton dimensions, as typed.", computed=False),
_field("units_per_carton", "Units per carton", "int", "Amazon units in one carton.",
computed=False, format={"decimals": 4}),
_field("carton_cbm", "Carton CBM", "int", "Cubic meters of one carton.",
computed=False, format={"decimals": 6}),
_field("amazon_cogs", "Amazon COGS", "currency", "Typed cost of one Amazon unit.", computed=False),
)
#: Owned columns Data Entry writes back besides the link (S82-T71). Their keys stay stable.
ENTRY_KEYS = ("carton_size", "units_per_carton", "carton_cbm", "amazon_cogs")
#: Tracking columns: a refresh writes them, people read them, and a period switch keeps them.
TRACK_FIELDS = (
_field("status_set", "Status set", "text", "The status and the day a refresh first saw it.",
computed=True),
_field("suggested_product", "Suggested Odoo product", "text",
f"The match a person can accept. It links {LINK_PURPOSE}.", computed=True),
_field("suggested_units", "Suggested units", "int", "Odoo units in one Amazon unit.",
computed=True, format={"decimals": 4}),
_field("match_rule", "Match rule", "text", "How the matcher found the product.", computed=True),
_field("match_rejected", "Rejected matches", "text", "Odoo codes a person rejected.",
computed=True),
# S82-T73 (R43): which cost a refresh used. A period switch keeps it.
_field("cogs_source", "COGS source", "text", "Odoo receipt, or Typed when no receipt exists.",
computed=True),
)
#: COGS source cells (R43): the Odoo receipt cost wins and a typed Amazon COGS fills SKUs with no receipt.
COGS_RECEIPT = "Odoo receipt"
COGS_TYPED = "Typed"
TYPED_COST_SOURCE = "Typed Amazon COGS"
COMPUTED_KEYS = tuple(f["key"] for f in COMPUTED_FIELDS)
OWNED_KEYS = tuple(f["key"] for f in OWNED_FIELDS)
TRACK_KEYS = tuple(f["key"] for f in TRACK_FIELDS)
ALL_KEYS = COMPUTED_KEYS + OWNED_KEYS + TRACK_KEYS
SUGGESTION_KEYS = ("suggested_product", "suggested_units", "match_rule")
METRIC_KEYS = tuple(column[0] for column in M.METRIC_COLUMNS)
#: A failed rollup tie blanks Amazon totals while independent receipt observations remain visible.
#: Metrics built from economics figures blank with them. Ads, stock, returns and listing
#: metrics come from their own mirror tables, so they stay.
_UNVERIFIED_KEYS = ("units_sold", "net_sales", "fees", "sponsored_ads", "net_proceeds",
"cogs_total", "net_profit", "margin", "band") + tuple(
k for k in ("tacos", "profit_health", "score", "priority", "units_30d", "net_sales_30d", "ad_spend",
"ad_spend_30d")
if k in METRIC_KEYS)
#: What a failed settlement component withholds (`withheld_parts`). Any failure also withholds profit.
_WITHHELD_BY_PART = {
"product sales": tuple(k for k in ("net_sales", "units_sold", "net_sales_30d", "units_30d", "tacos")
if k in ("net_sales", "units_sold") or k in METRIC_KEYS),
"order fees": ("fees",),
# Published fees is fees_total_amount, which carries the storage, reimbursement and account accruals.
"monthly storage": ("fees",),
"reimbursements": ("fees",),
"account fees": ("fees",),
"ads": tuple(k for k in ("sponsored_ads", "ad_spend", "ad_spend_30d", "tacos")
if k == "sponsored_ads" or k in METRIC_KEYS),
}
_WITHHELD_PROFIT = ("net_proceeds", "cogs_total", "net_profit", "margin", "band") + tuple(
k for k in ("profit_health", "score", "priority") if k in METRIC_KEYS)
#: A metric check that fails blanks these on every table. The ids and the notes stay.
_METRIC_VALUE_KEYS = tuple(k for k in METRIC_KEYS if k not in ("parent_asin",) + M.NOTE_KEYS)
#: Cost reasons (R5, D-11172 sibling): a blank COGS, profit or margin names what is missing.
LINK_NOTE = "Link this SKU to its Odoo product in Data Entry."
PRODUCT_MISSING_NOTE = "Linked Odoo product not found. Check the link in Data Entry."
GRAIN_LINK_NOTE = "Link each SKU here to its Odoo product in Data Entry."
GRAIN_COST_NOTE = "A SKU here has no cost. Check the SKU tab."
NO_COST_NOTE = "No Odoo cost for this SKU."
COST_FIGURES = ("cogs_per_unit", "cogs_total", "net_profit", "margin")
ODOO_CHANNEL_FIGURES = ("odoo_channel_sales", "odoo_channel_profit", "odoo_channel_margin")
#: The period, sales and rollup count figures. A blank one comes only from a refusal or an absent row.
SALES_FIGURES = ("period", "sku_count", "skus", "units_sold", "net_sales", "fees", "sponsored_ads", "net_proceeds")
#: Health labels, score, priority and trends. A blank one comes only from a refusal or an absent row.
HEALTH_FIGURES = ("stock_health", "profit_health", "ads_health", "listing_health", "score", "priority",
"units_30d", "net_sales_30d", "ad_spend_30d")
#: Every reason column and the figures it explains, on all three tables.
BLANK_REASONS = {"sales_note": SALES_FIGURES, **M.NOTE_GROUPS, "cost_note": COST_FIGURES,
"odoo_margin_source": ODOO_CHANNEL_FIGURES, "health_note": HEALTH_FIGURES}
#: Every declared figure that holds a number or a reason (S85-T125). A blank Band is "neither Losing nor Thin"
#: when net profit is set; otherwise the cost or sales note explains it.
DECLARED_FIGURES = tuple(dict.fromkeys(f for figures in BLANK_REASONS.values() for f in figures)) + ("band",)
BLANK_CHECK = "every blank says why"
#: Every refusal sentence starts so.
REFUSED_PREFIX = "Not verified"
def unexplained_blanks(row, keys=None):
"""`[figure key]` blank in `row` with an empty reason column. `keys`: the table's field keys.
A refusal in the Period cell is no reason: the rows route projects it to the period label."""
have = set(row if keys is None else keys)
out = [figure for note, figures in BLANK_REASONS.items() if not str(row.get(note) or "").strip()
for figure in figures if figure in have and row.get(figure) in (None, "")]
explained = any(str(row.get(k) or "").strip() for k in ("net_profit", "cost_note", "sales_note"))
if "band" in have and row.get("band") in (None, "") and not explained:
out.append("band")
return out
def declared_figures(keys):
"""The declared figures a table with these field keys holds, in declaration order."""
have = set(keys)
return [figure for figure in DECLARED_FIGURES if figure in have]
def row_active(row):
"""A row with units sold or net sales in its period."""
return M._active(row)
def blank_check(tables):
"""`every blank says why` over `{table key: (field keys, rows)}`: ok when no declared figure is blank
without its reason. Counts only; it names figures, never row keys or values."""
blanks = unexplained = rows = 0
figures = {}
for keys, table_rows_ in tables.values():
declared = declared_figures(keys)
for row in table_rows_:
rows += 1
blanks += sum(1 for figure in declared if row.get(figure) in (None, ""))
for figure in unexplained_blanks(row, keys):
unexplained += 1
figures[figure] = figures.get(figure, 0) + 1
return {"check": BLANK_CHECK, "ok": rows > 0 and not unexplained, "rows": rows, "blanks": blanks,
"unexplained": unexplained, "figures": dict(sorted(figures.items()))}
def cost_note(row, *, code=None, found=True):
"""Why a SKU's COGS, profit or margin is blank; "" when all four are set."""
blank = [k for k in COST_FIGURES if row.get(k) is None]
if not blank:
return ""
if not code and not found:
return LINK_NOTE
if not found:
return PRODUCT_MISSING_NOTE
if blank == ["margin"]:
return M.NO_NET_SALES_NOTE
source = str(row.get("cost_source") or "").strip()
if source and row.get("cogs_per_unit") is None:
return source if source.endswith(".") else source + "."
return NO_COST_NOTE
_IDENTITY_KEYS = ("sku", "picture", "asin", "parent_asin", "title", "projects", "projects_label", "history")
_PERIOD_KEYS = tuple(k for k in COMPUTED_KEYS if k not in _IDENTITY_KEYS + COMPETITOR_KEYS)
#: The profit block a rollup table carries; `_ECON_ADDITIVE` sums, margin and band recompute.
_ECON_ADDITIVE = ("units_sold", "net_sales", "fees", "sponsored_ads", "net_proceeds", "cogs_total",
"net_profit")
_ROLLUP_ECON = ("period",) + _ECON_ADDITIVE[:5] + ("sales_note",) + _ECON_ADDITIVE[5:] + ("margin", "cost_note", "band")
def _grain_fields(key_column, grain):
"""ASIN or parent ASIN table columns: ids, SKU count, the profit block, metrics, projects."""
econ = {f["key"]: f for f in _ECON_FIELDS}
metric = {f["key"]: f for f in _metric_fields()}
rollup = tuple(_field(key, label, ftype, description, computed=True, **dict(extra or {}))
for key, label, ftype, description, extra in M.ROLLUP_COLUMNS)
ids = ((econ["asin"], PICTURE_FIELD, metric["parent_asin"]) if key_column == "asin"
else (metric["parent_asin"], PICTURE_FIELD))
where = "this ASIN or its parent ASIN" if key_column == "asin" else "this parent ASIN"
progress = _progress_fields(f"Average progress of the projects on {where}.")
app = (_PDP_PROGRESS_FIELDS[:1] + _SIGNAL_FIELDS + _ASIN_DATE_FIELDS + (HISTORY_FIELD,)
if key_column == "asin" else (HISTORY_FIELD,))
companions = _ASIN_COMPANIONS if key_column == "asin" else ()
return (ids[:1] + progress[:1] + app + ids[1:] + (econ["title"],) + rollup + tuple(econ[k] for k in _ROLLUP_ECON)
+ _metric_fields(skip=("parent_asin",)) + (HEALTH_NOTE_FIELD,)
+ progress[1:] + companions)
ASIN_FIELDS = _grain_fields("asin", "asin")
PARENT_FIELDS = _grain_fields("parent_asin", "parent_asin")
_ROAS = next((c for c in M.METRIC_COLUMNS if c[0] == "roas"), ("roas", "ROAS", "int", "", {}))
#: Project columns people write: rows are added in the Grid and from the Projects popup.
PROJECT_USER_FIELDS = (
_field("title", "Title", "text", "What the project changes.", computed=False),
_field("grain", "Grain", "select", "SKU, ASIN or parent ASIN.", computed=False,
options=[M.GRAIN_LABELS[grain] for grain in M.GRAINS]),
_field("key", "Key", "text", "The SKU, ASIN or parent ASIN it works on.", computed=False),
_field("owner", "Owner", "user", "Who runs the project.", computed=False),
_field("start_on", "Start", "date", "When the work starts.", computed=False),
_field("deadline_on", "Deadline", "date", "When the work is due.", computed=False),
_field("completed_on", "Completed", "date", "When the work was done. Impact counts from this day.",
computed=False),
_field("status", "Status", "select", "Open or Done.", computed=False, options=list(PROJECT_STATUSES)),
)
#: Project columns a refresh writes.
PROJECT_COMPUTED_FIELDS = (
_field("health", "Health", "select", "On track, due this week, overdue or done, from the deadline.",
computed=True, options=list(PROJECT_HEALTH), format={"display": "dot"}),
_field("impact", "Impact", "currency", "Net sales since completion minus the same days one year earlier.",
computed=True),
_field("net_sales_since", "Net sales since", "currency", "Net sales from completion to yesterday.",
computed=True),
_field("net_sales_prior", "Net sales prior", "currency", "Net sales over the same days one year earlier.",
computed=True),
_field("units_since", "Units since", "int", "Units from completion to yesterday.", computed=True),
_field("units_prior", "Units prior", "int", "Units over the same days one year earlier.", computed=True),
_field("impact_units", "Impact units", "int", "Units since completion minus the same days one year earlier.",
computed=True),
_field("roas_since", "ROAS since", _ROAS[2], "Return on ad spend from completion to yesterday.",
computed=True, **dict(_ROAS[4] or {})),
_field("roas_prior", "ROAS prior", _ROAS[2], "Return on ad spend over the same days one year earlier.",
computed=True, **dict(_ROAS[4] or {})),
_field("impact_state", "Impact state", "text", "Why impact is blank.", computed=True),
)
PROJECT_FIELDS = PROJECT_USER_FIELDS + PROJECT_COMPUTED_FIELDS
PROJECT_USER_KEYS = tuple(f["key"] for f in PROJECT_USER_FIELDS)
PROJECT_COMPUTED_KEYS = tuple(f["key"] for f in PROJECT_COMPUTED_FIELDS)
def table_specs():
"""Every module table in creation order."""
specs = (
{"key": TABLE_KEY, "label": TABLE_LABEL, "fields": COMPUTED_FIELDS + OWNED_FIELDS + TRACK_FIELDS,
"owned": OWNED_KEYS, "keyColumn": "sku", "template": TEMPLATE_KEY},
{"key": ASIN_KEY, "label": ASIN_LABEL, "fields": ASIN_FIELDS + COMPETITOR_FIELDS, "owned": (),
"keyColumn": "asin", "template": ASIN_TEMPLATE_KEY},
{"key": PARENT_KEY, "label": PARENT_LABEL, "fields": PARENT_FIELDS + PARENT_COMPETITOR_FIELDS, "owned": (),
"keyColumn": "parent_asin", "template": PARENT_TEMPLATE_KEY},
)
for spec in specs:
fields = {f["key"]: f for f in spec["fields"]}
spec["fields"] = tuple(fields[key] for key in _project_field_order(fields))
return specs
# ---------------------------------------------------------------- period and threshold
def _today():
return dt.datetime.now(dt.timezone.utc).date()
def period_bounds(period=None, today=None):
"""`{key, start, end, label}` for 30d, 90d or YYYY-MM; ISO dates, end inclusive."""
key = str(period or PERIOD_DEFAULT).strip().lower()
today = today or _today()
if key in PERIOD_DAYS:
end = today - dt.timedelta(days=1)
start = end - dt.timedelta(days=PERIOD_DAYS[key] - 1)
label = f"Last {PERIOD_DAYS[key]} days"
else:
match = _MONTH.fullmatch(key)
if not match or not 1 <= int(match.group(2)) <= 12:
raise ValueError("Choose 30 days, 90 days or a calendar month.")
start = dt.date(int(match.group(1)), int(match.group(2)), 1)
if start > today:
raise ValueError("Choose a month that has started.")
following = (start.replace(day=28) + dt.timedelta(days=4)).replace(day=1)
end = following - dt.timedelta(days=1)
label = start.strftime("%B %Y")
return {"key": key, "start": start.isoformat(), "end": end.isoformat(),
"label": f"{label} ({start.isoformat()} to {end.isoformat()})"}
def period_choices(today=None, months=PICKER_MONTHS):
"""`[{key, label}]`: 30 days, 90 days, then the last `months` complete calendar months."""
out = [{"key": key, "label": f"Last {days} days"} for key, days in PERIOD_DAYS.items()]
first = (today or _today()).replace(day=1)
for _ in range(months):
first = (first - dt.timedelta(days=1)).replace(day=1)
out.append({"key": first.strftime("%Y-%m"), "label": first.strftime("%B %Y")})
return out
def clean_thin(value):
"""The Thin threshold as a fraction of net sales, 0 to under 1."""
if value is None or value == "":
return THIN_DEFAULT
try:
number = float(str(value).strip().rstrip("%"))
except (TypeError, ValueError):
raise ValueError("Enter the thin margin as a number.") from None
if str(value).strip().endswith("%"):
number /= 100.0
if not math.isfinite(number) or not 0 <= number < 1:
raise ValueError("Enter the thin margin between 0% and 99%.")
return number
def last_complete_month(today=None):
first = (today or _today()).replace(day=1)
return (first - dt.timedelta(days=1)).strftime("%Y-%m")
def verification_month(period=None, today=None):
"""The month a refresh ties to settlements: a picked month once it has ended, else the last one."""
key = str(period or PERIOD_DEFAULT).strip().lower()
last = last_complete_month(today)
return key if _MONTH.fullmatch(key) and key <= last else last
# ---------------------------------------------------------------- pure arithmetic
def _num(value):
try:
number = float(value)
except (TypeError, ValueError):
return 0.0
return number if math.isfinite(number) else 0.0
def band_for(net_profit, margin, thin=THIN_DEFAULT):
"""Losing below 0 profit, Thin below the threshold margin, else no band."""
if net_profit is None:
return None
if net_profit < 0:
return "Losing"
if margin is not None and margin < thin:
return "Thin"
return None
def compute_row(econ, cost=None, units_per_sku=1, thin=THIN_DEFAULT):
"""One SKU's figures. `cost` None means unlinked: COGS, profit, margin and band stay blank."""
net_sales = _num(econ.get("net_sales"))
fees = _num(econ.get("fees"))
ads = _num(econ.get("ads"))
units = int(_num(econ.get("units")))
net_proceeds = net_sales - fees - ads
row = {"sku": econ.get("sku"), "asin": econ.get("asin") or None,
"title": econ.get("title") or None, "units_sold": units, "net_sales": net_sales,
"fees": fees, "sponsored_ads": ads, "net_proceeds": net_proceeds,
"cogs_per_unit": None, "cogs_total": None, "net_profit": None, "margin": None,
"band": None}
if cost is None:
return row
cogs_per_unit = _num(cost) * (units_per_sku or 1)
cogs_total = cogs_per_unit * units
net_profit = net_proceeds - cogs_total
margin = (net_profit / net_sales) if net_sales else None
row.update(cogs_per_unit=cogs_per_unit, cogs_total=cogs_total, net_profit=net_profit,
margin=margin, band=band_for(net_profit, margin, thin))
return row
def cells(row, period_label=""):
"""Computed cells as stored strings; None means the cell is blank."""
out = {}
for key in _ECON_KEYS:
value = period_label if key == "period" else row.get(key)
if value is None or value == "":
out[key] = None
elif key in ("margin", "odoo_channel_margin"):
out[key] = f"{value * 100:.2f}"
elif key == "units_sold":
out[key] = str(int(value))
elif isinstance(value, float):
out[key] = f"{value:.2f}"
else:
out[key] = str(value)
return out
def odoo_code(value):
"""The Odoo internal reference a cell names: `RI-1` or `[RI-1] Name`."""
text = str(value or "").strip()
match = re.match(r"\[([^\]]+)\]", text)
return (match.group(1) if match else text).strip()
UNITS_DECIMALS = 4
def units_per_sku(value):
"""Odoo units in one Amazon unit, a positive number that may be a fraction (0.1 of a
1000-piece box). Blank, unreadable or not above 0 reads 1."""
try:
number = round(float(str(value).strip()), UNITS_DECIMALS)
except (TypeError, ValueError):
return 1.0
return number if math.isfinite(number) and number > 0 else 1.0
def units_text(value):
"""The stored cell for a units value: at most 4 decimals, no trailing zeros."""
return f"{units_per_sku(value):.{UNITS_DECIMALS}f}".rstrip("0").rstrip(".")
# ---------------------------------------------------------------- owned items (pure)
_STATUS_SET = re.compile(r"(.+) since (\d{4}-\d{2}-\d{2})")
def status_set_cell(status, day):
return f"{status} since {day.isoformat()}"
def parse_status_set(value):
"""`(status, date)` from a `status_set` cell, or `(None, None)`."""
match = _STATUS_SET.fullmatch(str(value or "").strip())
if not match:
return None, None
try:
return match.group(1), dt.date.fromisoformat(match.group(2))
except ValueError:
return None, None
def owned_changes(row, band, *, today, owner=None):
"""Cells a refresh writes on one SKU's owned item. It never overwrites a person's choice,
except the reopen: a fix 30 days old on a SKU still Losing or Thin goes back to Open."""
status = str(row.get("status") or "").strip()
seen, since = parse_status_set(row.get("status_set"))
due = (today + dt.timedelta(days=DUE_DAYS)).isoformat()
if not status:
if band in BANDS:
out = {"status": "Open", "status_set": status_set_cell("Open", today)}
if owner and not str(row.get("owner") or "").strip():
out["owner"] = owner
if not str(row.get("due") or "").strip():
out["due"] = due
return out
return {"status_set": None} if row.get("status_set") else {}
if seen != status or since is None:
return {"status_set": status_set_cell(status, today)}
if status in FIX_STATUSES and band in BANDS and (today - since).days >= REOPEN_DAYS:
return {"status": "Open", "due": due, "status_set": status_set_cell("Open", today)}
return {}
# ---------------------------------------------------------------- mirror reads (SQL only)
def _columns(con, table):
rows = con.execute("SELECT column_name FROM information_schema.columns WHERE table_name = ?",
[table]).fetchall()
return {r[0] for r in rows}
def _require(con, table, required):
cols = _columns(con, table)
if not cols:
raise SourceUnavailable(f"The Amazon mirror has no {table} table. Sync Amazon first.")
missing = [c for c in required if c not in cols]
if missing:
raise SourceUnavailable(f"{table} lacks {', '.join(missing)}. Resync Amazon economics.")
return cols
def ads_columns(cols):
"""Per ad type totals; charge detail columns are parts of these totals."""
return sorted(c for c in cols if _ADS_TOTAL.fullmatch(c) and "_charge_" not in c)
def economics_by_sku(con, start, end):
"""`[{sku, asin, units, net_sales, fees, ads}]`: every SKU in the mirror, period sums."""
cols = _require(con, ECONOMICS, _ECONOMICS_REQUIRED)
if "currency_code" in cols:
currencies = con.execute(f"SELECT count(DISTINCT currency_code) FROM {ECONOMICS}").fetchone()[0]
if currencies > 1:
raise SourceUnavailable("Amazon economics mixes currencies; profit needs one currency.")
ads = " + ".join(f'COALESCE("{c}", 0)' for c in ads_columns(cols)) or "0"
asin = "arg_max(child_asin, start_date)" if "child_asin" in cols else "NULL"
window = "FILTER (WHERE left(start_date, 10) BETWEEN ? AND ?)"
sql = (f"SELECT msku, {asin}, "
f"COALESCE(SUM(sales_net_units_sold) {window}, 0), "
f"COALESCE(SUM(sales_net_product_sales_amount) {window}, 0), "
f"COALESCE(SUM(fees_total_amount) {window}, 0), "
f"COALESCE(SUM({ads}) {window}, 0) "
f"FROM {ECONOMICS} WHERE NULLIF(trim(msku), '') IS NOT NULL GROUP BY msku ORDER BY msku")
rows = con.execute(sql, [start, end] * 4).fetchall()
return [{"sku": r[0], "asin": r[1], "units": r[2], "net_sales": r[3], "fees": r[4],
"ads": r[5]} for r in rows]
def titles(con):
"""`{seller_sku: item_name}` from the newest listing snapshot; empty without listings."""
cols = _columns(con, LISTINGS)
if not {"seller_sku", "item_name"} <= cols:
return {}
order = "COALESCE(_snapshot_date, '')" if "_snapshot_date" in cols else "1"
rows = con.execute(f"SELECT seller_sku, arg_max(item_name, {order}) FROM {LISTINGS} "
"WHERE seller_sku IS NOT NULL GROUP BY seller_sku").fetchall()
return {r[0]: r[1] for r in rows if r[1]}
def net_proceeds_total(con, start, end, *, drop_fees=False):
"""Net sales minus fees and ads over the window; `drop_fees` is the negative control."""
cols = _require(con, ECONOMICS, _ECONOMICS_REQUIRED)
ads = " + ".join(f'COALESCE("{c}", 0)' for c in ads_columns(cols)) or "0"
fees = "0" if drop_fees else "COALESCE(fees_total_amount, 0)"
return float(con.execute(
f"SELECT COALESCE(SUM(COALESCE(sales_net_product_sales_amount, 0) - {fees} - ({ads})), 0) "
f"FROM {ECONOMICS} WHERE left(start_date, 10) BETWEEN ? AND ?", [start, end]).fetchone()[0])
def reported_net_proceeds(con, start, end):
"""Amazon's own netProceeds total over the window, or None if the column is absent."""
if "net_proceeds_total_amount" not in _columns(con, ECONOMICS):
return None
return float(con.execute(
f"SELECT COALESCE(SUM(net_proceeds_total_amount), 0) FROM {ECONOMICS} "
"WHERE left(start_date, 10) BETWEEN ? AND ?", [start, end]).fetchone()[0])
def economics_currency(con):
if "currency_code" not in _columns(con, ECONOMICS):
return None
currency, count, missing = con.execute(
f"SELECT max(currency_code), count(DISTINCT currency_code), "
f"count(*) FILTER (WHERE NULLIF(trim(currency_code), '') IS NULL) FROM {ECONOMICS}").fetchone()
if count > 1 or missing:
raise SourceUnavailable("Amazon economics needs one known currency.")
return currency
def settlements_total(con, start, end, currency=None):
"""Sum posted settlements with both collected and withheld sales tax included.
`currency` keeps settlements whose header row names it: Amazon.ca and Amazon.com.mx settle
in their own currency, outside the economics marketplace. Their reports write DD.MM.YYYY.
"""
cols = _require(con, SETTLEMENTS, _SETTLEMENT_REQUIRED)
scoped = currency is not None and {"currency", "settlement_id"} <= cols
stream = ("JOIN (SELECT settlement_id FROM {t} WHERE currency IS NOT NULL GROUP BY settlement_id "
"HAVING any_value(currency) = ?) h USING (settlement_id)").format(t=SETTLEMENTS) if scoped else ""
total, rows, undated = con.execute(
"SELECT COALESCE(SUM(TRY_CAST(amount AS DECIMAL(38, 2))) "
"FILTER (WHERE d BETWEEN CAST(? AS DATE) AND CAST(? AS DATE)), 0), "
"count(*) FILTER (WHERE d BETWEEN CAST(? AS DATE) AND CAST(? AS DATE)), "
"count(*) FILTER (WHERE d IS NULL AND posted_date IS NOT NULL) "
"FROM (SELECT amount, posted_date, COALESCE(TRY_CAST(left(posted_date, 10) AS DATE), "
"CAST(try_strptime(left(posted_date, 10), '%d.%m.%Y') AS DATE)) AS d "
f"FROM {SETTLEMENTS} s {stream} WHERE COALESCE(transaction_type, '') <> ?)",
[start, end, start, end] + ([currency] if scoped else []) + [LOAN_REPAYMENT]).fetchone()
return {"total": float(total), "rows": int(rows), "undated": int(undated),
"currency": currency if scoped else None}
def _source_drill(con, sql, params, expected):
"""Read every private source row and return only its count, digest and replay query."""
digest, count = hashlib.sha256(), 0
cursor = con.execute(sql, params)
while rows := cursor.fetchmany(256):
for row in rows:
digest.update(json.dumps(row, default=str, ensure_ascii=True, separators=(",", ":")).encode())
digest.update(b"\n")
count += 1
return {"sql": sql, "params": params, "expectedRows": int(expected), "rows": count,
"countMatches": count == expected, "sha256": digest.hexdigest(), "executed": True}
def settlement_components(con, start, end, currency=None, *, drill=False):
"""Tie purchase-date order charges, service-month storage and daily ads to independent sources."""
month = dt.date.fromisoformat(start).strftime("%B %Y")
parts, problems, drills, named = [], [], [], []
def problem(component, reason):
problems.append(f"Not verified: {component} in {month}: {reason}.")
named.append(component)
def tie(component, ours, source, *, reason=None):
ours = float(ours)
source = None if source is None else float(source)
ok, gap = _tie(ours, source) if source is not None else (False, math.inf)
basis = {"product sales": "order purchase date", "order fees": "order purchase date",
"monthly storage": "storage service month", "ads": "daily ad report date"}
parts.append({"component": component, "month": start[:7], "ok": ok and not reason,
"a": ours, "b": source, "gap": gap, "tolerance": SETTLEMENT_TOLERANCE,
"dateBasis": basis.get(component, "order purchase date when present, otherwise posted date"),
"sourceTables": ([ECONOMICS] + [f"amzn_ads_{kind}_advertised_products" for kind in ("sp", "sb", "sd")]
if component == "ads" else [ECONOMICS, SETTLEMENTS, "amzn_orders"])})
if reason:
problem(component, reason)
elif not ok:
problem(component, f"independent Amazon amounts differ by {gap:.1%}")
sc = _columns(con, SETTLEMENTS)
ec = _columns(con, ECONOMICS)
required = set(_SETTLEMENT_REQUIRED) | {"settlement_id", "currency", "order_id", "amount_description"}
if not required <= sc:
problem("settlement components", "missing source columns " + ", ".join(sorted(required - sc)))
return {"total": None, "parts": parts, "reason": problems[0], "rows": 0}
if currency is None:
problem("settlement components", "economics currency unavailable")
return {"total": None, "parts": parts, "reason": problems[0], "rows": 0}
if not {"amazon_order_id", "purchase_date"} <= _columns(con, "amzn_orders"):
problem("order charges", "missing amzn_orders purchase dates")
return {"total": None, "parts": parts, "reason": problems[0], "rows": 0}
unscoped = con.execute(f"""SELECT count(*) FROM {SETTLEMENTS} s
LEFT JOIN (SELECT settlement_id FROM {SETTLEMENTS} GROUP BY settlement_id
HAVING count(DISTINCT currency) = 1) h USING (settlement_id)
WHERE h.settlement_id IS NULL AND (s.amount IS NOT NULL OR NULLIF(s.amount_type, '') IS NOT NULL)
AND (CAST({_moment_sql('posted_date')} AS DATE) BETWEEN CAST(? AS DATE) AND CAST(? AS DATE)
OR {_moment_sql('posted_date')} IS NULL)""", [start, end]).fetchone()[0]
if unscoped:
problem("settlement currency", "missing or conflicting settlement header currency")
# Aggregate in SQL before reading: duplicate order lines must not multiply settlement amounts.
cte = f"""WITH orders AS (
SELECT amazon_order_id, min(CAST({_moment_sql('purchase_date')} AS DATE)) AS bought,
count(DISTINCT CAST({_moment_sql('purchase_date')} AS DATE)) AS dates,
count(*) FILTER (WHERE {_moment_sql('purchase_date')} IS NULL) AS bad_dates
FROM amzn_orders GROUP BY amazon_order_id),
streams AS (SELECT settlement_id FROM {SETTLEMENTS} GROUP BY settlement_id
HAVING count(DISTINCT currency) = 1 AND max(currency) = ?),
lines AS (SELECT s.*, TRY_CAST(amount AS DECIMAL(38, 2)) AS currency_amount,
CAST({_moment_sql('s.posted_date')} AS DATE) AS posted,
o.bought, o.dates, o.bad_dates,
lower(regexp_replace(COALESCE(amount_description, ''), '[^a-zA-Z0-9]', '', 'g')) AS description,
lower(regexp_replace(COALESCE(amount_type, ''), '[^a-zA-Z0-9]', '', 'g')) AS kind,
lower(regexp_replace(COALESCE(transaction_type, ''), '[^a-zA-Z0-9]', '', 'g')) AS transaction
FROM {SETTLEMENTS} s JOIN streams USING (settlement_id)
LEFT JOIN orders o ON trim(s.order_id) = o.amazon_order_id
WHERE (amount IS NOT NULL OR NULLIF(amount_type, '') IS NOT NULL)
AND COALESCE(transaction_type, '') <> ?),
classified AS (SELECT *, CASE
WHEN concat(kind, description, transaction) LIKE '%inbounddefect%' THEN 'Inbound Defect Fee'
WHEN concat(kind, description, transaction) LIKE '%mcf%' THEN 'MCF credits'
WHEN kind = 'itemwithheldtax' OR description IN ('tax', 'shippingtax', 'giftwraptax')
OR description LIKE 'marketplacefacilitatortax%' THEN 'tax'
WHEN (kind = 'itemprice' AND (description IN ('shipping', 'giftwrap')
OR description LIKE '%promotion%')) OR kind IN ('promotion', 'shipping', 'shippingfees')
THEN 'shipping and promotions'
WHEN 'costofadvertising' IN (kind, description, transaction)
OR description IN ('advertising', 'advertisingfee') THEN 'ads billing'
WHEN description IN ('storagefee', 'fbamonthlystoragefee', 'monthlystoragefee')
OR 'fbainventorystoragefee' IN (kind, description, transaction) THEN 'monthly storage'
WHEN concat(kind, description, transaction) LIKE '%reimbursement%' THEN 'reimbursements'
WHEN description IN ('subscription', 'subscriptionfee', 'fbaremovalfee', 'fbadisposalfee',
'fbainboundtransportationfee', 'couponredemptionfee', 'dealfee')
OR kind IN ('dealfee', 'lightningdealfee', 'dealparticipationfee', 'dealperformancefee',
'fbacustomerreturnfee', 'fbacustomerreturnsfee', 'fbacustomerreturnperunitfee',
'fbacustomerreturnprocessingfee', 'fbainboundplacementservicefee')
OR description IN ('lightningdealfee', 'dealparticipationfee', 'dealperformancefee',
'fbacustomerreturnfee', 'fbacustomerreturnsfee', 'fbacustomerreturnperunitfee',
'fbacustomerreturnprocessingfee', 'fbainboundplacementservicefee')
OR transaction IN ('dealfee', 'lightningdealfee', 'dealparticipationfee', 'dealperformancefee',
'fbacustomerreturnfee', 'fbacustomerreturnsfee', 'fbacustomerreturnperunitfee',
'fbacustomerreturnprocessingfee', 'fbainboundplacementservicefee') THEN 'account fees'
WHEN NULLIF(trim(order_id), '') IS NOT NULL AND kind = 'itemprice'
AND description = 'principal' THEN 'product sales'
WHEN NULLIF(trim(order_id), '') IS NOT NULL AND kind IN ('itemfee', 'itemfees') THEN 'order fees'
ELSE 'unmatched component' END AS component FROM lines),
dated AS (SELECT *, CASE
WHEN component = 'monthly storage' THEN CAST(date_trunc('month', posted) - INTERVAL '1 month' AS DATE)
WHEN NULLIF(trim(order_id), '') IS NOT NULL THEN bought
ELSE posted END AS basis FROM classified)
"""
window = """basis BETWEEN CAST(? AS DATE) AND CAST(? AS DATE)
OR ((basis IS NULL OR dates <> 1 OR bad_dates <> 0)
AND (posted BETWEEN CAST(? AS DATE) AND CAST(? AS DATE) OR posted IS NULL))"""
source_params = [currency, LOAN_REPAYMENT, start, end, start, end]
groups = con.execute(cte + """SELECT component, COALESCE(SUM(currency_amount), 0), count(*),
count(*) FILTER (WHERE currency_amount IS NULL OR posted IS NULL),
count(*) FILTER (WHERE NULLIF(trim(order_id), '') IS NOT NULL
AND (bought IS NULL OR dates <> 1 OR bad_dates <> 0))
FROM dated WHERE """ + window + " GROUP BY component ORDER BY component", source_params).fetchall()
amounts = {name: total for name, total, *_ in groups}
if not groups:
problem("settlement components", "no settlement source rows on the purchase or service dates")
if drill:
counts = {row[0]: row[2] for row in groups}
for name in sorted(set(counts) | {"product sales", "order fees", "monthly storage",
"reimbursements", "account fees", "tax"}):
query = (cte + "SELECT settlement_id, order_id, posted_date, amount_type, amount_description, "
"amount, bought, dates, bad_dates, component, basis FROM dated WHERE (" + window
+ ") AND component = ? ORDER BY ALL")
drills.append(dict(_source_drill(con, query, source_params + [name], counts.get(name, 0)),
component="sales tax" if name == "tax" else name, source=SETTLEMENTS))
for name, _total, _n, bad, unmatched in groups:
if bad or unmatched:
problem(name, "unmatched purchase date, unreadable posting date or unavailable amount")
if name in ("unmatched component", "Inbound Defect Fee", "MCF credits"):
problem(name, "settlement category has no economics mapping")
fee_columns = sorted(c for c in ec if c.startswith("fee_") and c.endswith("_amount") and "_charge_" not in c)
storage_columns = [c for c in fee_columns if "storage" in c]
reimbursement_columns = [c for c in fee_columns if "reimbursement" in c]
account_columns = [c for c in fee_columns if any(name in c for name in
("subscription", "removal", "disposal", "inbound_transportation", "inbound_placement",
"coupon", "deal_", "customer_return", "return_processing", "returns_processing"))]
# Storage fee columns are accruals in economics; only monthly settlement charges use the prior month.
sums = ["sales_net_product_sales_amount", "fees_total_amount", *ads_columns(ec),
*storage_columns, *reimbursement_columns, *account_columns]
observed = con.execute("SELECT " + ", ".join(f'COALESCE(SUM(TRY_CAST("{c}" AS DECIMAL(38, 2))), 0)' for c in sums)
+ ", count(*) FILTER (WHERE "
+ " OR ".join(f'("{c}" IS NOT NULL AND TRY_CAST("{c}" AS DECIMAL(38, 2)) IS NULL)'
for c in sums) + ")"
+ f" FROM {ECONOMICS} WHERE left(start_date, 10) BETWEEN ? AND ?",
[start, end]).fetchone()
if observed[-1]:
problem("settlement components", "unavailable economics amount")
values = dict(zip(sums, observed[:-1]))
sales, fees = values[sums[0]], values[sums[1]]
storage = sum(values[c] for c in storage_columns)
reimbursements = sum(values[c] for c in reimbursement_columns)
account_fees = sum(values[c] for c in account_columns)
ads = sum(values[c] for c in ads_columns(ec))
if drill:
def expression(columns):
return " + ".join(f'COALESCE(TRY_CAST("{c}" AS DECIMAL(38, 2)), 0)' for c in columns) or "0"
non_order = storage_columns + reimbursement_columns + account_columns
component_columns = {"product sales": ["sales_net_product_sales_amount"],
"order fees": ["fees_total_amount", *non_order],
"monthly storage": storage_columns, "reimbursements": reimbursement_columns,
"account fees": account_columns, "ads": ads_columns(ec)}
for name, columns in component_columns.items():
amount = expression(columns)
if name == "order fees":
amount = expression(["fees_total_amount"]) + " - (" + expression(non_order) + ")"
invalid = " OR ".join(f'("{c}" IS NOT NULL AND TRY_CAST("{c}" AS DECIMAL(38, 2)) IS NULL)'
for c in columns) or "FALSE"
where = f"left(start_date, 10) BETWEEN ? AND ? AND (({amount}) <> 0 OR {invalid})"
count = con.execute(f"SELECT count(*) FROM {ECONOMICS} WHERE {where}", [start, end]).fetchone()[0]
selected = list(dict.fromkeys(["start_date", "msku", "currency_code", *columns]))
projection = ", ".join(f'"{c}"' for c in selected if c in ec)
query = f"SELECT {projection} FROM {ECONOMICS} WHERE {where} ORDER BY ALL"
drills.append(dict(_source_drill(con, query, [start, end], count), component=name, source=ECONOMICS))
tie("product sales", sales, amounts.get("product sales", 0.0))
tie("order fees", fees - storage - reimbursements - account_fees, -amounts.get("order fees", 0.0))
tie("monthly storage", storage, -amounts.get("monthly storage", 0.0))
tie("reimbursements", reimbursements, -amounts.get("reimbursements", 0.0))
tie("account fees", account_fees, -amounts.get("account fees", 0.0))
tie("sales tax", 0.0, amounts.get("tax", 0.0))
# Shipping and promotions lie outside net product sales; ad invoices use a separate billing period.
if ads or "ads billing" in amounts:
ad_total = 0
for kind in ("sp", "sb", "sd"):
table = f"amzn_ads_{kind}_advertised_products"
if not {"date", "cost"} <= _columns(con, table):
problem("ads", f"missing source {table}.cost")
ad_total = None
break
total, count, present = con.execute(
f"SELECT SUM(TRY_CAST(cost AS DECIMAL(38, 2))), count(*), "
f"count(TRY_CAST(cost AS DECIMAL(38, 2))) FROM {table} "
"WHERE left(CAST(date AS VARCHAR), 10) BETWEEN ? AND ?", [start, end]).fetchone()
if drill:
query = (f"SELECT date, cost FROM {table} WHERE left(CAST(date AS VARCHAR), 10) "
"BETWEEN ? AND ? ORDER BY ALL")
drills.append(dict(_source_drill(con, query, [start, end], count), component="ads", source=table))
if not count or count != present or total is None or not math.isfinite(float(total)):
problem("ads", f"unavailable daily amounts in {table}")
ad_total = None
break
ad_total += total
tie("ads", ads, ad_total)
else:
ad_total = 0
target = (float(sum(amounts.get(name, 0) for name in
("product sales", "order fees", "monthly storage", "reimbursements", "account fees")))
- float(ad_total)) if ad_total is not None else None
return {"total": target, "parts": parts, "reason": problems[0] if problems else None,
**({"drills": drills, "problems": problems} if drill else {}),
"failedComponents": list(dict.fromkeys(named)),
"rows": sum(row[2] for row in groups),
"excluded": {k: float(amounts[k]) for k in ("shipping and promotions", "ads billing") if k in amounts}}
_COVERAGE_REQUIRED = ("settlement_id", "total_amount", "settlement_start_date", "settlement_end_date")
def _moment_sql(name):
"""A settlement header moment as a UTC timestamp; Amazon.ca and .com.mx write DD.MM.YYYY."""
text = f"replace(trim(CAST({name} AS VARCHAR)), ' UTC', '')"
return (f"COALESCE(TRY_CAST({text} AS TIMESTAMPTZ) AT TIME ZONE 'UTC', "
f"try_strptime({text}, ['%d.%m.%Y %H:%M:%S', '%d.%m.%Y %H:%M', '%d.%m.%Y']))")
def _day_label(day):
return f"{day.day} {day:%b}"
def settlement_coverage(con, start, end, currency=None):
"""The days of `start`..`end` inside issued settlements, cut at the last full day, and the clear day.
Without settlement header dates in the mirror the whole window is the tie range, as before.
`start` is None when no issued settlement covers a full day of the window. `clear` is the last
close plus the median settlement cycle, stepped until it passes `end`.
"""
lo, hi = dt.date.fromisoformat(start), dt.date.fromisoformat(end)
out = {"start": start, "end": end, "partial": False, "clear": None, "source": "window"}
cols = _columns(con, SETTLEMENTS)
if not set(_COVERAGE_REQUIRED) <= cols:
return _coverage_words(out)
scoped = currency is not None and "currency" in cols
rows = con.execute(
f"SELECT {_moment_sql('settlement_start_date')}, {_moment_sql('settlement_end_date')} "
f"FROM {SETTLEMENTS} WHERE total_amount IS NOT NULL"
+ (" AND trim(CAST(currency AS VARCHAR)) = ?" if scoped else ""),
[currency] if scoped else []).fetchall()
spans = sorted((s, e) for s, e in rows if s is not None and e is not None and e > s)
if not spans:
return _coverage_words(out)
merged = []
for s, e in spans:
if merged and s <= merged[-1][1]:
merged[-1][1] = max(merged[-1][1], e)
else:
merged.append([s, e])
midnight = dt.time(0, 0)
covered = None
for s, e in merged:
# A close is mid-day, so its own day is partial: only whole days count on both sides.
first = s.date() if s.time() == midnight else s.date() + dt.timedelta(days=1)
last = (e + dt.timedelta(seconds=1)).date() - dt.timedelta(days=1)
first, last = max(first, lo), min(last, hi)
if first <= last:
covered = (first, last)
out["source"] = "settlements"
if covered is None:
out.update(start=None, end=None)
return _coverage_words(out)
out.update(start=covered[0].isoformat(), end=covered[1].isoformat(),
partial=covered != (lo, hi))
if covered[1] < hi:
cycles = sorted(e - s for s, e in spans)
cycle = cycles[len(cycles) // 2]
close = max(e for _s, e in spans)
for _ in range(60):
if (close + dt.timedelta(seconds=1)).date() - dt.timedelta(days=1) >= hi:
break
close += cycle
out["clear"] = close.date().isoformat()
return _coverage_words(out)
def _coverage_words(out):
"""Adds `label` and `sentence`: the verified range and the day the rest is due to settle."""
if out["start"] is None:
return dict(out, label=None, sentence=None)
label = " to ".join(_day_label(dt.date.fromisoformat(out[k])) for k in ("start", "end"))
sentence = f"Verified {label}."
if out["clear"]:
sentence += f" The rest settles about {_day_label(dt.date.fromisoformat(out['clear']))}."
return dict(out, label=label, sentence=sentence)
def _newest_close(con, currency=None):
"""The newest issued settlement close in `currency`, or None without settlement header dates."""
cols = _columns(con, SETTLEMENTS)
if not set(_COVERAGE_REQUIRED) <= cols:
return None
scoped = currency is not None and "currency" in cols
close = con.execute(
f"SELECT max({_moment_sql('settlement_end_date')}) FROM {SETTLEMENTS} WHERE total_amount IS NOT NULL"
+ (" AND trim(CAST(currency AS VARCHAR)) = ?" if scoped else ""),
[currency] if scoped else []).fetchone()[0]
return close if isinstance(close, dt.datetime) else None
def _month_end(month):
first = dt.date.fromisoformat(f"{month}-01")
return (first.replace(day=28) + dt.timedelta(days=4)).replace(day=1) - dt.timedelta(days=1)
def settled_month(con, month, currency=None):
"""`(month, settling)`: `month` stepped back, at most twice, while Amazon still holds its funds.
A month is settling while the newest close is before its last day plus SETTLEMENT_DEFERRAL_DAYS.
`settling` is None when `month` is settled or the mirror has no settlement header dates."""
close = _newest_close(con, currency)
if close is None:
return month, None
asked = month
for _ in range(2):
if close.date() >= _month_end(month) + dt.timedelta(days=SETTLEMENT_DEFERRAL_DAYS):
break
month = (dt.date.fromisoformat(f"{month}-01") - dt.timedelta(days=1)).strftime("%Y-%m")
if month == asked:
return month, None
end = _month_end(asked)
due = end + dt.timedelta(days=SETTLEMENT_DEFERRAL_DAYS)
clear = settlement_coverage(con, f"{asked}-01", end.isoformat(), currency).get("clear")
if clear and dt.date.fromisoformat(clear) > due:
due = dt.date.fromisoformat(clear)
words = {m: dt.date.fromisoformat(f"{m}-01").strftime("%B %Y") for m in (asked, month)}
line = (f"{words[asked]} is still settling, so the figures tie to {words[month]}. "
f"{words[asked]} settles about {_day_label(due)}.")
return month, {"month": asked, "verifiedMonth": month, "clear": due.isoformat(), "line": line}
# ---------------------------------------------------------------- Odoo cost (read only)
def odoo_costs(codes):
"""`({code: {id, name, cost}}, problems)` for linked references; search_read only (rule 4)."""
codes = sorted({c for c in codes if c})
if not codes:
return {}, []
import core.odoo as O
domain = [("default_code", "in", codes), ("active", "in", [True, False])]
rows = O.search_read("product.product", domain,
["id", "default_code", "name", "standard_price"], limit=len(codes) * 4 + 50)
expected = O.get_odoo().search_count("product.product", domain)
if len(rows) != expected:
raise SourceUnavailable(f"Odoo returned {len(rows)} of {expected} linked products; "
"the cost read is incomplete.")
found, problems = {}, []
by_code = {}
for row in rows:
by_code.setdefault(str(row.get("default_code") or ""), []).append(row)
for code in codes:
matches = by_code.get(code) or []
if len(matches) == 1:
row = matches[0]
found[code] = {"id": row["id"], "name": row.get("name") or "",
"cost": _num(row.get("standard_price"))}
elif matches:
problems.append(f"{code}: {len(matches)} Odoo products share this reference")
else:
problems.append(f"{code}: no Odoo product has this reference")
return found, problems
def bound_cost_reader(runtime):
"""Legacy explicit cost injection, bound to the tenant and unused by production refreshes."""
def read(codes):
source = runtime.odoo_source() if hasattr(runtime, "odoo_source") else None
if source is None:
return {}, [f"{code}: this workspace has no Odoo connection" for code in sorted(codes)]
with source._bound():
return odoo_costs(codes)
return read
# ---------------------------------------------------------------- receipt costs (mirror only)
_RECEIPT_COLUMNS = {
"stock_valuation_layer": {"product_id", "quantity", "value", "create_date", "stock_move_id"},
"stock_move": {"id", "product_id", "date", "state", "picking_id", "location_id", "location_dest_id"},
"stock_location": {"id", "usage"},
"stock_picking": {"id", "name", "date_done"},
"stock_valuation_adjustment_lines": {"id", "cost_id", "product_id", "move_id", "quantity",
"additional_landed_cost", "write_date"},
"stock_landed_cost": {"id", "date", "state", "amount_total"},
}
_RECEIPT_FROM = (
"stock_valuation_layer v JOIN stock_move m ON m.id=v.stock_move_id AND m.product_id=v.product_id "
"JOIN stock_location src ON src.id=m.location_id "
"JOIN stock_location dst ON dst.id=m.location_dest_id "
"LEFT JOIN stock_picking p ON p.id=m.picking_id")
_RECEIPT_DATE = ("left(COALESCE(NULLIF(CAST(p.date_done AS VARCHAR), ''), "
"NULLIF(CAST(m.date AS VARCHAR), ''), CAST(v.create_date AS VARCHAR)), 10)")
def _require_odoo(con, required):
"""Require physical mirror columns and completed population and column checkpoints."""
tracked = {"entity", "phase"} <= _columns(con, "_sync_state")
for table, fields in required.items():
if not fields <= _columns(con, table):
raise SourceUnavailable(f"{table}: receipt or margin source unavailable. Sync Odoo.")
if not tracked:
raise SourceUnavailable(f"{table}: source completeness unavailable. Sync Odoo.")
if tracked:
from harness import datastore as DS
key = DS._connection_state_key(table)
state = con.execute("SELECT phase FROM _sync_state WHERE entity=?", [key]).fetchone()
if not state or state[0] != "live":
raise SourceUnavailable(f"{table}: source coverage incomplete. Sync Odoo.")
spec = DS.ENTITIES.get(table) or {}
fields_to_check = set(spec.get("backfill_fields") or ()) - {"write_date"}
if table == "stock_landed_cost":
fields_to_check.update(("date", "state", "amount_total"))
if fields_to_check:
states = DS._column_backfill_states(con, table)
if any(states.get(field, (None,))[0] != "done" for field in fields_to_check):
raise SourceUnavailable(f"{table}: source columns incomplete. Sync Odoo.")
def mirror_products(con, codes):
"""Resolve unique internal references without reading standard_price."""
if not codes:
return {}, []
try:
_require_odoo(con, {"product_product": {"id", "default_code", "name"}})
except SourceUnavailable as exc:
return {}, [str(exc)]
codes = sorted(codes)
rows = con.execute("SELECT id, default_code, name FROM product_product WHERE default_code IN ("
+ ",".join("?" for _ in codes) + ") ORDER BY id", codes).fetchall()
grouped, found, problems = {}, {}, []
for pid, code, name in rows:
grouped.setdefault(code, []).append({"id": int(pid), "name": name})
for code in codes:
candidates = grouped.get(code, [])
if len(candidates) == 1:
found[code] = candidates[0]
else:
problems.append(f"{code}: Odoo product reference missing or ambiguous")
return found, problems
def receipt_costs(con, product_ids=None):
"""Read vendor receipts and posted adjustments, excluding zero-quantity revaluations."""
if product_ids is not None and not product_ids:
return []
_require_odoo(con, _RECEIPT_COLUMNS)
params = sorted(set(product_ids or ()))
scope = (" AND v.product_id IN (" + ",".join("?" for _ in params) + ")") if params else ""
rows = con.execute(
f"SELECT v.product_id, m.id, p.name, {_RECEIPT_DATE} AS day, SUM(v.quantity), SUM(v.value), "
f"count(*) FILTER (WHERE v.value IS NULL OR NOT isfinite(v.value)) FROM {_RECEIPT_FROM} "
"WHERE m.state='done' AND src.usage='supplier' AND dst.usage='internal' AND v.quantity>0"
+ scope + " GROUP BY v.product_id, m.id, p.name, day ORDER BY day, m.id, v.product_id", params).fetchall()
adjustments = con.execute(
"SELECT a.id, a.cost_id, a.product_id, a.move_id, a.quantity, a.additional_landed_cost, "
"CAST(a.write_date AS VARCHAR), h.state, CAST(h.date AS VARCHAR) "
"FROM stock_valuation_adjustment_lines a LEFT JOIN stock_landed_cost h ON h.id=a.cost_id"
+ (" WHERE a.product_id IN (" + ",".join("?" for _ in params) + ")" if params else "")
+ " ORDER BY a.id", params).fetchall()
by_move = {}
for aid, cid, pid, move, quantity, amount, written, state, day in adjustments:
if state is None:
raise SourceUnavailable("Landed cost header missing. Sync Odoo before using receipt costs.")
if state != "done":
continue
by_move.setdefault((pid, move), []).append({"id": int(aid), "cost_id": int(cid),
"quantity": None if quantity is None else float(quantity),
"amount": None if amount is None else float(amount), "write_date": written, "date": day})
receipts = []
for pid, move, name, day, quantity, value, invalid in rows:
lines = by_move.get((pid, move), [])
invalid_line = any(line["quantity"] is None or _num(line["quantity"]) <= 0
or line["amount"] is None or not math.isfinite(float(line["amount"]))
or not line["date"] for line in lines)
if invalid or invalid_line or not name or not day or not math.isfinite(float(quantity)):
raise SourceUnavailable("Receipt valuation or landed allocation incomplete. Sync Odoo.")
try:
dt.date.fromisoformat(day)
except ValueError:
raise SourceUnavailable("Receipt date unavailable. Sync Odoo.") from None
purchase = float(value) / float(quantity)
landed = sum(float(line["amount"]) / float(line["quantity"]) for line in lines)
row = {"receipt_key": f"{int(pid)}:{int(move)}", "product_id": int(pid), "move_id": int(move),
"receipt": name, "date": day, "quantity": float(quantity),
"purchase_unit_cost": purchase, "landed_unit_cost": landed,
"unit_cost": purchase + landed, "landed_lines": lines,
"source": "Landed cost" if lines else "Purchase price (no landed costs)"}
row["observation_id"] = hashlib.sha256(json.dumps(row, sort_keys=True).encode()).hexdigest()
receipts.append(row)
return receipts
def read_cost_history(st):
"""Read saved observations strictly so a failed read cannot replace history."""
# A TenantRuntime has no _read_strict of its own: read through its backend's, under its key.
reader = getattr(st, "_read_strict", None)
if reader is None and callable(getattr(st, "_store", None)) and callable(getattr(st, "store_key", None)):
strict = getattr(st._store(), "_read_strict", None)
if callable(strict):
reader = lambda name: strict(st.store_key(name)) # noqa: E731
doc = (reader(COST_HISTORY_KEY) if reader else st.get(COST_HISTORY_KEY, fresh=True)) or {}
required = {"observation_id", "product_id", "move_id", "receipt", "date", "quantity",
"purchase_unit_cost", "landed_unit_cost", "unit_cost", "source"}
if (not isinstance(doc, dict) or doc.get("version", 1) != 1
or not isinstance(doc.get("receipts", []), list)
or any(not isinstance(row, dict) or not required <= row.keys() for row in doc.get("receipts", []))):
raise TableRefused("Receipt cost history is unreadable. Restore it before refreshing.")
return doc.get("receipts", [])
def append_cost_history(st, receipts):
"""Append new receipt evidence and retain every previous observation unchanged."""
stamp = dt.datetime.now(dt.timezone.utc).isoformat(timespec="microseconds")
def append(cur):
cur = dict(cur or {})
prior = cur.get("receipts", [])
if not isinstance(prior, list) or any(not isinstance(row, dict) or not row.get("observation_id")
for row in prior):
raise TableRefused("Receipt cost history is unreadable. Restore it before refreshing.")
seen = {row["observation_id"] for row in prior}
added = []
for row in receipts:
if row["observation_id"] not in seen:
added.append(dict(row, observed_at=stamp))
seen.add(row["observation_id"])
return {**cur, "version": 1, "receipts": prior + added}
st.update(COST_HISTORY_KEY, append, flush="sync")
return read_cost_history(st)
def _history_text(history, product_id):
return "\n".join(
f"{row['receipt']} | {row['date']} | {row['quantity']:g} Odoo units | "
f"Purchase {row['purchase_unit_cost']:.2f} | Landed {row['landed_unit_cost']:.2f} | "
f"Total {row['unit_cost']:.2f} per Odoo unit | {row['source']}"
for row in history if row["product_id"] == product_id)
def odoo_channel_margins(con, start, end):
"""Read confirmed margins of the tenant's wholesale teams, excluding GIFTWARE DEALS."""
required = {"sale_order_line": {"order_id", "product_id", "margin", "price_subtotal"},
"sale_order": {"id", "state", "team_id", "partner_id", "date_order"},
"res_partner": {"id", "name"}}
try:
_require_odoo(con, required)
except SourceUnavailable:
return {}, "Odoo margin unavailable"
import core.odoo as O
teams = O.active_team_ids()
if not teams:
return {}, "Odoo margin unavailable"
rows = con.execute(
"SELECT l.product_id, SUM(l.margin), SUM(l.price_subtotal), "
"count(*) FILTER (WHERE l.margin IS NULL OR l.price_subtotal IS NULL "
"OR NOT isfinite(l.margin) OR NOT isfinite(l.price_subtotal)) "
"FROM sale_order_line l JOIN sale_order o ON o.id=l.order_id "
"JOIN res_partner p ON p.id=o.partner_id WHERE o.state IN ('sale','done') "
f"AND o.team_id IN ({','.join('?' * len(teams))}) "
"AND upper(p.name) NOT LIKE '%GIFTWARE DEALS%' "
"AND l.product_id IS NOT NULL "
"AND left(CAST(o.date_order AS VARCHAR),10) BETWEEN ? AND ? GROUP BY l.product_id",
[*teams, start, end]).fetchall()
return {int(pid): {"profit": None if missing else float(profit),
"sales": None if missing else float(sales),
"margin": float(profit) / float(sales) if sales and not missing else None,
"source": "Odoo margin unavailable" if missing else "Confirmed Fisch and Royal sales"}
for pid, profit, sales, missing in rows}, "Confirmed Fisch and Royal sales"
def fifo_costs(con, start, end, row_links, products, receipts, history):
"""Allocate each product once by receipt date, sale date and seller SKU, including prior depletion."""
saved = {row["observation_id"]: row for row in history}
pools = {}
for receipt in receipts:
observed = saved.get(receipt["observation_id"])
if observed is not None:
pools.setdefault(receipt["product_id"], []).append({"receipt": observed,
"left": receipt["quantity"]})
linked = {sku: (products[link["code"]]["id"], units_per_sku(link.get("units_per_sku")))
for sku, link in row_links.items() if link.get("code") in products}
if not linked:
return {}, []
product_ids = sorted({pid for pid, _multiplier in linked.values()})
seller_skus = sorted(linked)
events = con.execute(
f"SELECT left(CAST(start_date AS VARCHAR),10), msku, SUM(sales_net_units_sold) "
f"FROM {ECONOMICS} WHERE left(CAST(start_date AS VARCHAR),10)<=? "
"AND msku IN (" + ",".join("?" for _ in seller_skus) + ") "
"GROUP BY 1,2 ORDER BY 1,2", [end, *seller_skus]).fetchall()
outgoing = con.execute(
"SELECT v.product_id, left(CAST(m.date AS VARCHAR),10) AS day, m.id, -SUM(v.quantity) "
f"FROM {_RECEIPT_FROM} WHERE m.state='done' AND src.usage='internal' "
"AND dst.usage<>'internal' AND v.quantity<0 AND left(CAST(m.date AS VARCHAR),10)<? "
"AND v.product_id IN (" + ",".join("?" for _ in product_ids) + ") "
"GROUP BY v.product_id, day, m.id ORDER BY day,m.id", [start, *product_ids]).fetchall()
prior_products = {int(row[0]) for row in outgoing}
incomplete = set()
usage, result = [], {}
def consume(pid, quantity, day, sku=None):
remaining, used = quantity, []
for batch in pools.get(pid, []):
receipt = batch["receipt"]
if remaining <= 1e-8:
break
if receipt["date"] > day or batch["left"] <= 1e-8:
continue
taken = min(batch["left"], remaining)
batch["left"] -= taken
remaining -= taken
used.append({"sku": sku, "product_id": pid, "move_id": receipt["move_id"],
"sale_date": day,
"observation_id": receipt["observation_id"], "units": taken,
"purchase_used": taken * receipt["purchase_unit_cost"],
"landed_used": taken * receipt["landed_unit_cost"], "receipt": receipt})
return used, max(remaining, 0.0)
for pid, day, _move, quantity in outgoing:
_used, missing = consume(int(pid), float(quantity), day)
if missing > 1e-8:
incomplete.add(int(pid))
prior_amazon, prior_outgoing = {}, {}
for day, sku, units in events:
if day and day < start and sku in linked:
pid, multiplier = linked[sku]
prior_amazon[pid] = prior_amazon.get(pid, 0.0) + max(float(units or 0), 0.0) * multiplier
for pid, _day, _move, quantity in outgoing:
prior_outgoing[int(pid)] = prior_outgoing.get(int(pid), 0.0) + float(quantity)
incomplete.update(pid for pid in prior_products if prior_outgoing[pid] + 1e-8 < prior_amazon.get(pid, 0.0))
for day, sku, units in events:
if sku not in linked or not day:
continue
pid, multiplier = linked[sku]
quantity = float(units or 0) * multiplier
if day < start and pid in prior_products:
continue
if quantity < 0:
if day < start:
incomplete.add(pid)
else:
result.setdefault(sku, {"units": 0.0, "missing": 0.0, "usage": []})["returns"] = True
continue
used, missing = consume(pid, quantity, day, sku if day >= start else None)
if day < start:
if missing > 1e-8:
incomplete.add(pid)
continue
row = result.setdefault(sku, {"units": 0.0, "missing": 0.0, "usage": []})
row["units"] += quantity
row["missing"] += missing
row["usage"].extend(used)
usage.extend(used)
for sku, (pid, _multiplier) in linked.items():
row = result.setdefault(sku, {"units": 0.0, "missing": 0.0, "usage": []})
row["incomplete_prior"] = pid in incomplete
row["prior_source"] = "Odoo stock moves" if pid in prior_products else "Amazon unit history"
return result, usage
def _odoo_binding(runtime):
source = runtime.odoo_source() if hasattr(runtime, "odoo_source") else None
if source is None:
raise SourceUnavailable("This workspace has no Odoo connection.")
return source._bound()
# ---------------------------------------------------------------- suggested matches (R13)
#: Rule labels a `match_rule` cell stores, in the order the matcher tries them.
RULE_CODE = "Odoo code"
RULE_PREFIX = "Odoo code without prefix"
RULE_PACK = "Odoo code for the pack"
RULE_AMAZON = "Amazon Odoo code"
RULE_BARCODE = "Barcode"
RULE_NAME = "Name"
RULES = (RULE_CODE, RULE_PREFIX, RULE_PACK, RULE_AMAZON, RULE_BARCODE, RULE_NAME)
#: Listing report product id types that carry a barcode: 3 is UPC, 4 is EAN.
BARCODE_ID_TYPES = ("3", "4")
_NAME_STOP = frozenset("a an and the of for with per in x h to by or on inch inches set pack "
"piece pieces box bag roll count pcs".split())
_PACK = re.compile(r"(.+?)_(\d{1,4})")
_SELLER_PREFIX = re.compile(r"[A-Za-z]{2,3}[-_ ](?=.)")
_ODOO_PACK = re.compile(r"(\d+)\s*-?\s*pieces?\s+per\s+(?:box|pack|case|bag)", re.I)
_TITLE_PACK = re.compile(r"\b(?:set|pack)\s+of\s+(\d+)\b|\b(\d+)\s*-?\s*(?:pack|pcs|pieces|count)\b", re.I)
def title_pack(title):
"""Pieces in one Amazon unit as the listing title states them ("Set of 12", "100 Pack"), or None."""
match = _TITLE_PACK.search(str(title or ""))
n = int(match.group(1) or match.group(2)) if match else 0
return n if n > 0 else None
def odoo_pack(name):
"""Pieces in one Odoo unit from its name (`24-Piece Per Box` is 24), or None."""
match = _ODOO_PACK.search(str(name or ""))
return int(match.group(1)) if match and int(match.group(1)) > 0 else None
def odoo_units(n, name):
"""Odoo units in one Amazon unit of `n` pieces: n / k when the Odoo name states k pieces,
to 4 decimals and never 0; n when it states none."""
n = max(int(n or 1), 1)
k = odoo_pack(name)
if k is None:
return float(n)
return max(round(n / k, UNITS_DECIMALS), 10.0 ** -UNITS_DECIMALS)
def _norm_code(value):
return re.sub(r"[^A-Z0-9]", "", str(value or "").upper())
def _norm_barcode(value):
return re.sub(r"\D", "", str(value or "")).lstrip("0")
def _name_tokens(text, brand=()):
text = str(text or "").lower().split("|")[0]
return {w for w in re.findall(r"[a-z0-9]+(?:\.[0-9]+)?", text)
if w not in _NAME_STOP and w not in brand and (len(w) > 1 or w.isdigit())}
def odoo_catalogue():
"""Every Odoo product, active and archived, in one read; search_read and search_count only."""
import core.odoo as O
domain = [("active", "in", [True, False])]
expected = O.get_odoo().search_count("product.product", domain)
rows = O.search_read("product.product", domain,
["id", "default_code", "barcode", "name"], limit=expected + 50)
if len(rows) != expected:
raise SourceUnavailable(f"Odoo returned {len(rows)} of {expected} products; "
"the product read is incomplete.")
return rows
def listing_barcodes(con):
"""`{seller_sku: barcode}` from the newest listing snapshot, UPC and EAN ids only."""
cols = _columns(con, LISTINGS)
if not {"seller_sku", "product_id", "product_id_type"} <= cols:
return {}
order = "COALESCE(_snapshot_date, '')" if "_snapshot_date" in cols else "1"
rows = con.execute(f"SELECT seller_sku, arg_max(product_id, {order}), "
f"arg_max(CAST(TRY_CAST(product_id_type AS BIGINT) AS VARCHAR), {order}) FROM {LISTINGS} "
"WHERE seller_sku IS NOT NULL GROUP BY seller_sku").fetchall()
return {r[0]: r[1] for r in rows if r[1] and str(r[2]).strip() in BARCODE_ID_TYPES}
class Matcher:
"""Suggests one Odoo product per Amazon SKU. Only a product whose code is unique can link."""
def __init__(self, products, brand=("royal", "imports")):
codes = {}
for p in products:
code = str(p.get("default_code") or "").strip()
if code and code != "False":
codes.setdefault(code, []).append(p)
self.by_code = {code: rows[0] for code, rows in codes.items() if len(rows) == 1}
self.by_norm, self.by_barcode = {}, {}
for code, p in self.by_code.items():
self.by_norm.setdefault(_norm_code(code), []).append(code)
barcode = p.get("barcode")
if barcode and barcode is not False and _norm_barcode(barcode):
self.by_barcode.setdefault(_norm_barcode(barcode), []).append(code)
self.brand = frozenset(brand)
self.tokens = {code: _name_tokens(p.get("name"), self.brand) for code, p in self.by_code.items()}
df = {}
for toks in self.tokens.values():
for w in toks:
df[w] = df.get(w, 0) + 1
total = max(len(self.tokens), 1)
self.idf = {w: math.log(total / (1 + n)) + 1.0 for w, n in df.items()}
self.index = {}
for code, toks in self.tokens.items():
for w in toks:
self.index.setdefault(w, []).append(code)
def label(self, code):
return f"[{code}] {self.by_code[code].get('name') or ''}".strip()
def _one(self, codes, rejected):
codes = [c for c in codes or () if c not in rejected]
return codes[0] if len(codes) == 1 else None
def _by_code(self, candidate, rejected):
exact = [c for c in self.by_norm.get(_norm_code(candidate), ()) if c.upper() == candidate.upper()]
return self._one(exact, rejected) or self._one(self.by_norm.get(_norm_code(candidate)), rejected)
def code_candidates(self, sku):
"""`[(candidate code, pieces, rule)]` in the order the code rules try them. Pieces is
the `_n` count for the pack stem, 0 for a pack code (one Odoo unit), None otherwise."""
raw = str(sku or "").strip()
out = [(raw, None, RULE_CODE)]
plain = re.sub(r"\s*\(#?[^)]*\)\s*$", "", raw).lstrip(".#-_ ")
base = _SELLER_PREFIX.sub("", plain, count=1)
out += [(plain, None, RULE_PREFIX), (base, None, RULE_PREFIX)]
pack = _PACK.fullmatch(base)
if pack:
stem, n = pack.group(1), int(pack.group(2))
out += [(f"{stem}-{n:02d}", 0, RULE_PACK), (stem, max(n, 1), RULE_PACK)]
out += [("AMZ" + c, u, RULE_AMAZON) for c, u, _ in list(out)]
return [item for item in out if item[0]]
def pack_units(self, sku):
base = _SELLER_PREFIX.sub("", str(sku or "").strip().lstrip(".#-_ "), count=1)
pack = _PACK.fullmatch(re.sub(r"\s*\(#?[^)]*\)\s*$", "", base))
return max(int(pack.group(2)), 1) if pack else None
def by_name(self, title, rejected):
"""The product whose every name word (three or more) is in the title, with a clear lead
on the next. A short generic name such as `WEDDING ARCH` matches too many titles."""
words = _name_tokens(title, self.brand)
if not words:
return None
seen, scored = set(), []
for w in words:
for code in self.index.get(w, ()):
if code in seen or code in rejected:
continue
seen.add(code)
toks = self.tokens[code]
if len(toks) >= 3 and sum(not t[0].isdigit() for t in toks) >= 2 and toks <= words:
scored.append((sum(self.idf[t] for t in toks), code))
scored.sort(reverse=True)
if not scored or (len(scored) > 1 and scored[0][0] - scored[1][0] < 1.0):
return None
return scored[0][1]
def _units(self, code, n):
"""Pieces in the Amazon unit become Odoo units; 0 (a pack code) or unknown stays 1."""
return odoo_units(n, self.by_code[code].get("name")) if n else 1.0
def suggest(self, sku, *, title=None, barcode=None, rejected=()):
"""`{code, product, units, rule}` or None."""
rejected = set(rejected or ())
suffix = self.pack_units(sku)
pieces = suffix or title_pack(title)
for candidate, n, rule in self.code_candidates(sku):
code = self._by_code(candidate, rejected)
if code:
if n is None:
n = None if suffix else pieces
return {"code": code, "product": self.label(code), "rule": rule,
"units": self._units(code, n)}
if barcode and _norm_barcode(barcode):
code = self._one(self.by_barcode.get(_norm_barcode(barcode)), rejected)
if code:
return {"code": code, "product": self.label(code), "units": self._units(code, pieces),
"rule": RULE_BARCODE}
code = self.by_name(title, rejected) if title else None
if code:
return {"code": code, "product": self.label(code), "units": self._units(code, pieces),
"rule": RULE_NAME}
return None
def _rejected_codes(row):
return [c for c in (s.strip() for s in str(row.get("match_rejected") or "").split(",")) if c]
def decision_changes(row, state):
"""Cells a person's Linked or Wrong match writes on one row; {} when there is nothing to apply."""
state = LEGACY_MATCH_STATES.get(state, state)
if str(row.get("odoo_product") or "").strip():
return {}
suggested = str(row.get("suggested_product") or "").strip()
if state == CONFIRMED and suggested:
return {"odoo_product": suggested,
"units_per_sku": units_text(row.get("suggested_units")),
"match_state": CONFIRMED}
if state == REJECTED and suggested:
rejected = _rejected_codes(row)
code = odoo_code(suggested)
if code and code not in rejected:
rejected.append(code)
return {"match_state": REJECTED, "match_rejected": ", ".join(rejected),
**{k: None for k in SUGGESTION_KEYS}}
return {}
def _apply_cells(row, changes):
for key, value in changes.items():
if value is None:
row.pop(key, None)
else:
row[key] = value
def apply_decisions(st, *, by=WRITER, sku=None, state=None):
"""Apply every Linked or Wrong match status a person set, or one door's decision on
`sku`. Returns `{confirmed, rejected}` and, for a door, the row after the change."""
import core.user_tables as ut
counts, after = {}, {}
def _set(cur):
counts.clear()
after.clear()
counts.update(confirmed=0, rejected=0)
table = (cur or {}).get(TABLE_KEY)
if table is None:
return cur
for row in (table.get("rows") or {}).values():
row_sku = str(row.get("sku") or "").strip()
if sku is not None and row_sku != sku:
continue
want = state if sku is not None else str(row.get("match_state") or "").strip()
want = LEGACY_MATCH_STATES.get(want, want)
changes = decision_changes(row, want)
if changes:
_apply_cells(row, changes)
counts["confirmed" if want == CONFIRMED else "rejected"] += 1
if sku is not None:
after.update(row)
return cur
ut.update_with_activity(TABLE_KEY, _set, st=st, by=by or WRITER, via=VIA, row_ids=None)
return {**counts, "row": dict(after) if sku is not None else None}
def suggest_matches(st, con, products, *, by=WRITER):
"""Suggest an Odoo product for every unlinked SKU, no cap, in one store write. A SKU whose
Odoo product is set is never touched. Suggestions go to this table only, never to Odoo."""
import core.user_tables as ut
matcher = Matcher(products)
names, barcodes = titles(con), listing_barcodes(con)
counts = {}
def _set(cur):
counts.clear()
counts.update(considered=0, suggested=0, unmatched=0, **{rule: 0 for rule in RULES})
table = (cur or {}).get(TABLE_KEY)
if table is None:
return cur
for row in (table.get("rows") or {}).values():
sku = str(row.get("sku") or "").strip()
if not sku or str(row.get("odoo_product") or "").strip():
continue
counts["considered"] += 1
state = str(row.get("match_state") or "").strip()
found = matcher.suggest(sku, title=names.get(sku) or row.get("title"),
barcode=barcodes.get(sku), rejected=_rejected_codes(row))
if found is None:
counts["unmatched"] += 1
changes = {k: None for k in SUGGESTION_KEYS}
if state in (NEEDS_REVIEW, CONFIRMED):
changes["match_state"] = None
else:
counts["suggested"] += 1
counts[found["rule"]] += 1
changes = {"suggested_product": found["product"], "match_state": NEEDS_REVIEW,
"suggested_units": units_text(found["units"]), "match_rule": found["rule"]}
_apply_cells(row, {k: v for k, v in changes.items()
if v != row.get(k) and not (v is None and k not in row)})
return cur
ut.update_with_activity(TABLE_KEY, _set, st=st, by=by or WRITER, via=VIA, row_ids=None)
return dict(counts)
def _log(line):
try:
print(f"amazon_profit: {line}", flush=True)
except Exception: # noqa: BLE001
pass
def _sync_cursor(tenant_key, rt):
"""`(cursor, close)` on the tenant mirror the sync just wrote, else a fresh stored snapshot."""
from harness import amazon_store
path = Path(amazon_store._tenant_mirror_path(tenant_key, rt))
if path.is_file():
from harness import datastore
try:
shared = Path(datastore.DB_PATH).resolve() == path.resolve()
except Exception: # noqa: BLE001
shared = False
if shared:
cur = datastore.ro_cursor()
else:
import duckdb
cur = duckdb.connect(str(path), read_only=True)
if _mirror_has_economics(cur) or {'sku', 'attributes', 'summaries'} <= _columns(cur, 'amzn_listing_items'):
return cur, cur.close
cur.close()
for key in [k for k in _SNAPSHOT if k[0] == tenant_key]:
_SNAPSHOT.pop(key, None)
cur = snapshot_cursor(tenant_key)
return cur, cur.close
def sync_matches(tenant_key, synced_tables, *, st=None, con=None, catalogue=None, today=None):
"""Refresh the current period, apply decisions, then suggest for every unlinked SKU.
Returns the report. It raises on failure; `on_amazon_sync` is the never-raise wrapper."""
synced = {str(t or "").strip().lower() for t in synced_tables or ()}
if not synced & {"economics", "listings", "listing_items"}:
return {"state": "skipped", "synced": sorted(synced)}
if st is None:
from harness import amazon_store
st = amazon_store._runtime_for(tenant_key)
if st is None:
raise SourceUnavailable(f"The store for {tenant_key} could not be opened.")
close = None
if con is None:
con, close = _sync_cursor(tenant_key, st)
try:
report = {"state": "done"}
if synced & {"listings", "listing_items"}:
try:
report["partDates"] = _part_dates.observe(st, con, day=today or _today())
report["partDateCells"] = paint_asin_part_dates(st)
except Exception as exc: # quality-ok: Date storage failure must not block profit matching.
report['partDatesError'] = f'{type(exc).__name__}: {exc}'
_log(f"{tenant_key}: listing part dates failed ({report['partDatesError']}).")
if not synced & {'economics', 'listings'}:
return dict(report, state='failed', reason=report['partDatesError'])
if not synced & {"economics", "listings"}:
return report
try:
refreshed = refresh(st, con, by=WRITER, today=today, tenant=tenant_key)
report["refresh"] = {k: refreshed.get(k) for k in ("skus", "linked", "refusal")}
except Exception as exc: # noqa: BLE001
report["refreshError"] = f"{type(exc).__name__}: {exc}"
_log(f"{tenant_key}: the profit refresh failed ({report['refreshError']}); "
"suggesting on the rows already stored.")
with _LOCK:
ensure_table(st, WRITER)
report["decisions"] = apply_decisions(st)
if catalogue is None:
with _odoo_binding(st):
products = odoo_catalogue()
else:
products = catalogue()
with _LOCK:
report["matches"] = suggest_matches(st, con, products)
write_settings(st, matchedAt=dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds"),
matches=report["matches"], matchError=None)
return report
finally:
if close is not None:
try:
close()
except Exception: # noqa: BLE001
pass
def on_amazon_sync(tenant_key, synced_tables):
"""Contract C2 callback after an Amazon sync. Never raises: a failure logs one line naming
the cause, links nothing and returns."""
try:
return sync_matches(tenant_key, synced_tables)
except Exception as exc: # noqa: BLE001
reason = f"{type(exc).__name__}: {exc}"
_log(f"{tenant_key}: no Odoo matches suggested; every SKU stays unlinked. Cause: {reason}")
try:
from harness import amazon_store
st = amazon_store._runtime_for(tenant_key)
if st is not None:
write_settings(st, matchError=reason)
except Exception: # noqa: BLE001
pass
return {"state": "failed", "reason": reason}
# ---------------------------------------------------------------- module-owned table
def _definition_fields(fields=None, owned_keys=None):
"""Stored field definitions for one module table; the SKU table by default."""
import core.user_tables as ut
fields = tuple(COMPUTED_FIELDS + OWNED_FIELDS + TRACK_FIELDS if fields is None else fields)
owned = set(OWNED_KEYS if owned_keys is None else owned_keys)
keys = _project_field_order([f["key"] for f in fields])
by_key = {}
# One field per call: `clean_fields` slices a list at `MAX_FIELDS`.
for raw in fields:
for field in ut.clean_fields([dict(raw)]) or ():
by_key[field["key"]] = field
missing = [key for key in keys if key not in by_key]
if missing:
raise TableRefused(f"The Command Center columns failed the table field check: {', '.join(missing)}.")
out = []
for key in keys:
field = dict(by_key[key])
# editRole governs the column definition only. Cell values on owned columns follow the
# database edit grant (`may_edit_database`), so editors edit them and viewers cannot.
field["editRole"] = "admins"
if key not in owned:
field["automation"] = {"preset": True, "flowId": VIA}
out.append(field)
return out
def _project_field_order(keys):
"""Pin App fields in schema defaults and keep companions last, without changing saved layouts."""
keys = list(keys)
if not keys or keys[0] not in ("sku", "asin", "parent_asin"):
return keys
pinned = ASIN_APP_KEYS if keys[0] == "asin" else (keys[0], "projects", "history")
companions = (("projects_label",) + tuple(f["key"] for f in _ASIN_COMPANIONS)
if keys[0] == "asin" else ())
return ([key for key in pinned if key in keys]
+ [key for key in keys if key not in pinned + companions]
+ [key for key in companions if key in keys])
def _format_repairs(fields, wanted):
"""`{key: format}` for module fields an adopted table holds without their format."""
want = {f["key"]: f["format"] for f in wanted if f.get("format")}
return {f.get("key"): want[f.get("key")] for f in fields or ()
if isinstance(f, dict) and f.get("key") in want and not f.get("format")}
def _grain_word_repairs(fields, wanted):
"""`{key: {attr: value}}` where a stored table still shows grain keys instead of grain words."""
want = {f["key"]: f for f in wanted}
out = {}
for field in fields or ():
target = want.get(field.get("key")) if isinstance(field, dict) else None
if target is None:
continue
if field["key"] == "grain" and target.get("options") and field.get("options") != target["options"]:
out.setdefault("grain", {})["options"] = list(target["options"])
if field["key"] in ("projects", "pdp_update") and (field.get("type") != target["type"] or "link" in field
or (field.get("format") or {}) != (target.get("format") or {})):
out[field["key"]] = {"type": target["type"], "format": dict(target.get("format") or {}), "link": None}
return out
#: SKU fields whose label, description or options changed. A stored copy is rewritten once.
_RELABELLED = ("odoo_product", "match_state", "suggested_product")
def _relabel_repairs(fields):
"""`{key: {attr: value}}` where a stored SKU table still holds an old label, description or option."""
want = {f["key"]: f for f in OWNED_FIELDS + TRACK_FIELDS if f["key"] in _RELABELLED}
out = {}
for field in fields or ():
target = want.get(field.get("key")) if isinstance(field, dict) else None
if target is None:
continue
fix = {attr: target[attr] for attr in ("label", "description") if field.get(attr) != target[attr]}
if target.get("options"):
# Options a person added stay after the wanted ones; legacy words become the new ones.
extra = [o for o in field.get("options") or () if o not in target["options"]
and o not in LEGACY_MATCH_STATES]
options = list(target["options"]) + extra
if field.get("options") != options:
fix["options"] = options
if fix:
out[field["key"]] = fix
return out
def _legacy_cells(rows):
"""Row ids whose Link status cell still holds an old option word."""
return [rid for rid, row in (rows or {}).items()
if isinstance(row, dict) and row.get("match_state") in LEGACY_MATCH_STATES]
def _relabel(table):
"""The SKU table repair step: new labels, descriptions and options, and every stored
Confirmed or Rejected cell rewritten to Linked or Wrong match. Keys never change."""
repairs = _relabel_repairs(table.get("fields"))
for field in table.get("fields") or []:
field.update(repairs.get(field.get("key")) or {})
rows = table.get("rows") or {}
for rid in _legacy_cells(rows):
rows[rid]["match_state"] = LEGACY_MATCH_STATES[rows[rid]["match_state"]]
#: The core helper's field placement, kept under this name for its gate.
_insert_missing = module_tables._insert_missing
def ensure_table(st, username, *, table_key=None, label=None, fields=None, owned_keys=None):
"""Create or adopt module tables through `core.module_tables`. Returns the key.
With no `table_key` it ensures every table in `table_specs` order and returns `TABLE_KEY`.
Format, grain word, cost description and Link status (`_relabel`) repairs stay here."""
if table_key is None:
# module_tables persists only `fieldLimit`; user_tables.field_limit reads that key.
specs = [{"key": s["key"], "label": s["label"], "fields": _definition_fields(s["fields"], s["owned"]),
"fieldLimit": PROFIT_MAX_FIELDS}
for s in table_specs()]
else:
specs = [{"key": table_key, "label": label or table_key, "fields": _definition_fields(fields, owned_keys)}]
cost_description = next((field["description"] for field in _ECON_FIELDS if field["key"] == "cogs_per_unit"))
def _stale(spec, have):
wanted = spec["fields"]
field_keys = [field["key"] for field in have.get("fields") or []]
if _project_field_order(field_keys) != field_keys:
return True
if _format_repairs(have.get("fields"), wanted) or _grain_word_repairs(have.get("fields"), wanted):
return True
if spec["key"] == TABLE_KEY and (_relabel_repairs(have.get("fields")) or _legacy_cells(have.get("rows"))):
return True
return spec["key"] == TABLE_KEY and any(f.get("key") == "cogs_per_unit" and f.get("description") != cost_description
for f in have.get("fields") or [])
def _repair(spec, table):
repairs = _format_repairs(table.get("fields"), spec["fields"])
words = _grain_word_repairs(table.get("fields"), spec["fields"])
for field in table.get("fields") or []:
if field.get("key") in repairs:
field["format"] = dict(repairs[field["key"]])
field.update(words.get(field.get("key")) or {})
if "link" in field and field["link"] is None:
field.pop("link")
if spec["key"] == TABLE_KEY and field.get("key") == "cogs_per_unit":
field["description"] = cost_description
if spec["key"] == TABLE_KEY:
_relabel(table)
fields_by_key = {field["key"]: field for field in table.get("fields") or []}
table["fields"] = [fields_by_key[key] for key in _project_field_order(fields_by_key)]
keys = module_tables.ensure_module_table(st, username, specs, owner=WRITER, stale=_stale, repair=_repair,
error=TableRefused)
return keys[0] if table_key is not None else TABLE_KEY
def _ensure_table_grants(st, keys):
"""Add the everyone view grant to each stored module table that lacks it; keep other entries."""
module_tables.ensure_everyone_grant(st, keys, owner=WRITER)
def table_rows(st, table_key=TABLE_KEY):
import core.user_tables as ut
return dict((ut.get_one(table_key, st) or {}).get("rows") or {})
def typed_costs(rows):
"""`{sku: typed Amazon COGS}` for every SKU row whose typed value is a number of 0 or more."""
out = {}
for row in rows.values():
sku = str(row.get("sku") or "").strip()
text = str(row.get("amazon_cogs") or "").strip().replace(",", "").lstrip("$")
try:
value = float(text)
except ValueError:
continue
if sku and math.isfinite(value) and value >= 0:
out[sku] = value
return out
def links(rows):
"""`{sku: {code, units_per_sku}}` from the owned columns."""
out = {}
for row in rows.values():
sku = str(row.get("sku") or "").strip()
if sku:
out[sku] = {"code": odoo_code(row.get("odoo_product")),
"units_per_sku": units_per_sku(row.get("units_per_sku"))}
return out
def _rid_order(rid):
text = str(rid)
return (0, int(text), text) if text.isdigit() else (1, 0, text)
def merge_duplicates(st):
"""Keep one row per SKU (R10): the oldest row takes empty owned cells from its copies,
then the copies go. Returns how many rows were removed."""
import core.user_tables as ut
groups = {}
for rid, row in table_rows(st).items():
sku = str(row.get("sku") or "").strip()
if sku:
groups.setdefault(sku, []).append(str(rid))
extra = {}
for rids in groups.values():
if len(rids) > 1:
rids.sort(key=_rid_order)
extra[rids[0]] = rids[1:]
if not extra:
return 0
dropped = [rid for rids in extra.values() for rid in rids]
def _merge(cur):
table = cur.get(TABLE_KEY)
if table is None:
return cur
rows = table.get("rows") or {}
order = table.get("_rowCreationOrder") or {}
for keep, copies in extra.items():
row = rows.get(keep)
if row is None:
continue
for rid in copies:
copy = rows.pop(rid, None) or {}
order.pop(rid, None)
for key in OWNED_KEYS + TRACK_KEYS:
if not str(row.get(key) or "").strip() and str(copy.get(key) or "").strip():
row[key] = copy[key]
return cur
ut.update_with_activity(TABLE_KEY, _merge, st=st, by=WRITER, via=VIA,
row_ids=list(extra) + dropped, flush="sync")
return len(dropped)
def write_rows(st, computed, *, table_key=TABLE_KEY, key_column="sku", label=TABLE_LABEL, period_keys=None):
"""Upsert computed cells by the key column through `core.module_tables`; owned cells are never touched.
A stored row absent from `computed` keeps its row and loses its period figures."""
period_keys = _PERIOD_KEYS if period_keys is None else tuple(period_keys)
report = module_tables.write_module_rows(st, table_key, computed, key_column=key_column, writer=WRITER, via=VIA,
label=label, clear_keys=period_keys, origin="Amazon", error=TableRefused)
reason = "No Amazon economics for this row. Sync Amazon and refresh."
wanted = {rid: {note: reason for note, figures in BLANK_REASONS.items()
if set(figures) & set(period_keys) and row.get(note) != reason}
for rid, row in table_rows(st, table_key).items()
if str(row.get(key_column) or "").strip() not in computed}
wanted = {rid: values for rid, values in wanted.items() if values}
for rid, row in table_rows(st, table_key).items():
if row.get("history") != "View":
wanted.setdefault(rid, {})["history"] = "View"
if wanted:
import core.user_tables as ut
def _notes(cur):
rows = ((cur or {}).get(table_key) or {}).get("rows") or {}
for rid, values in wanted.items():
if rid in rows:
rows[rid].update(values)
return cur
ut.update_with_activity(table_key, _notes, st=st, by=WRITER, via=VIA, row_ids=list(wanted))
return report
# ---------------------------------------------------------------- settings
def read_settings(st):
raw = {}
try:
raw = st.get(SETTINGS_KEY) or {}
except Exception: # noqa: BLE001
raw = {}
raw = raw if isinstance(raw, dict) else {}
period = raw.get("period") or PERIOD_DEFAULT
try:
period_bounds(period)
except ValueError:
period = PERIOD_DEFAULT
try:
thin = clean_thin(raw.get("thin"))
except ValueError:
thin = THIN_DEFAULT
owner = str(raw.get("defaultOwner") or "").strip().lower() or None
return {**raw, "period": period, "thin": thin, "defaultOwner": owner}
def write_settings(st, **values):
def _up(cur):
cur = dict(cur) if isinstance(cur, dict) else {}
cur.update(values)
return cur
st.update(SETTINGS_KEY, _up, flush="sync")
# ---------------------------------------------------------------- default owner (R31)
#: Accounts that act for the platform, never for the business.
SERVICE_ACCOUNTS = ("admin", "automation", "scheduler", "qa-runner", "leadership")
def _accounts(tenant):
"""`[(username, record)]`: the tenant's active accounts, through the one tenant filter."""
from core import users
return users._tenant_accounts(tenant)
def _is_service(username, record):
name = str(username or "").strip().lower()
return (not name or name.startswith("qa") or name in SERVICE_ACCOUNTS
or name.endswith("-admin") or bool(record.get("platform_admin")))
def owner_account(tenant=TENANT_ZERO):
"""The tenant's owner account: no record names one, so the earliest active Administrator
that is a person. A person signs in with an email; earliest is the oldest password stamp."""
people = [(name, rec) for name, rec in _accounts(tenant)
if isinstance(rec, dict) and rec.get("role") == "admin" and not _is_service(name, rec)]
if not people:
return None
people.sort(key=lambda item: ("@" not in item[0], not item[1].get("password_changed_at"),
str(item[1].get("password_changed_at") or ""), item[0]))
return str(people[0][0]).strip().lower()
def account_name(username, tenant=TENANT_ZERO):
"""The display name a `user` cell stores for `username`, or None if no active account."""
want = str(username or "").strip().lower()
for name, rec in _accounts(tenant):
if str(name).strip().lower() == want and isinstance(rec, dict):
return str(rec.get("name") or name)
return None
def set_default_owner(st, username, tenant=TENANT_ZERO, by=""):
"""Point new owned items at `username`; refuses an account this tenant does not hold."""
want = str(username or "").strip().lower()
if not want or account_name(want, tenant) is None:
raise ValueError("Choose an active account in this workspace.")
write_settings(st, defaultOwner=want, defaultOwnerSetBy=by or WRITER)
return want
def default_owner(st, tenant=TENANT_ZERO):
"""`(username, display name)` new owned items get; resolves and stores it on first use."""
username = read_settings(st)["defaultOwner"]
if not username:
username = owner_account(tenant)
if username:
write_settings(st, defaultOwner=username, defaultOwnerSetBy=WRITER)
return username, (account_name(username, tenant) if username else None)
# ---------------------------------------------------------------- seeded views
def seeded_views(template_key=TEMPLATE_KEY):
from core import view_templates as vt
return list((vt.get(template_key) or {}).get("views") or ())
def ensure_views(st, owner, *, table_key=TABLE_KEY, template_key=TEMPLATE_KEY):
"""Write the seeded views once, owned by `owner`, granted to everyone who can open the table.
A view lives in one personal stratum and reaches others through a `*` view grant, the same
residency every shared view has. A view a person deleted stays deleted: `viewsSeeded` in the
settings records what was written."""
from core import shares
from core import view_templates as vt
from core.table_store import TableStore
done = set(read_settings(st).get("viewsSeeded") or ())
ops = TableStore(vt.workspace_key(table_key), st)
owner = str(owner or WRITER).strip().lower()
made = []
for view in seeded_views(template_key):
if view["id"] in done:
continue
found = ops.find_view(view["id"])
holder = found[0] if found else owner
if not found:
ops.save_view(owner, dict(view, createdBy=owner), actor=WRITER, origin={"client": VIA})
made.append(view["id"])
held = shares.grants("view", view["id"], st=st)
if not any(e["user"] == shares.EVERYONE for e in held["entries"]):
shares.set_grants("view", view["id"],
held["entries"] + [{"user": shares.EVERYONE, "role": "view"}],
owner=held["owner"] or holder, st=st)
done.add(view["id"])
added = set(read_settings(st).get("viewColumnsAdded") or ())
for view in seeded_views(template_key):
filter_mark = f"{view['id']}:verified-cost-filter"
if view["id"].endswith("_cogs_missing") and filter_mark not in added:
found = ops.find_view(view["id"])
config = dict((found[1].get("config") or {}) if found else {})
legacy = [{"colId": "odoo_product", "op": "isEmpty"},
{"colId": "cogs_per_unit", "op": "isEmpty"}]
if found and config.get("filters") == legacy and config.get("filterConj") == "or":
config.update(filters=list(view["config"]["filters"]), filterConj=view["config"]["filterConj"])
ops.save_view(found[0], dict(found[1], config=config), actor=WRITER, origin={"client": VIA})
added.add(filter_mark)
for column in _VIEW_COLUMNS_ADDED:
mark = f"{view['id']}:{column}"
if mark in added or column not in view["config"]["visible"]:
continue
found = None if view["id"] in made else ops.find_view(view["id"])
config = dict((found[1].get("config") or {}) if found else {})
if found and column not in [*(config.get("order") or ()), *(config.get("visible") or ())]:
before = _column_before(view["config"]["visible"], column)
config.update({part: _insert_after(config[part], before, column) for part in ("order", "visible")
if isinstance(config.get(part), list)})
ops.save_view(found[0], dict(found[1], config=config), actor=WRITER, origin={"client": VIA})
added.add(mark)
for column in _VIEW_COLUMNS_DROPPED:
mark = f"{view['id']}:-{column}"
if mark in added or column in view["config"]["visible"]:
continue
found = None if view["id"] in made else ops.find_view(view["id"])
config = dict((found[1].get("config") or {}) if found else {})
if found and column in (config.get("visible") or ()):
config["visible"] = [c for c in config["visible"] if c != column]
ops.save_view(found[0], dict(found[1], config=config), actor=WRITER, origin={"client": VIA})
added.add(mark)
write_settings(st, viewsSeeded=sorted(done), viewColumnsAdded=sorted(added))
return made
#: Columns a seeded view already written gains once; a person who hides one later keeps it hidden.
#: The S85-T125 metric columns follow template order, so each lands after the column before it.
_VIEW_COLUMNS_ADDED = ("picture", "odoo_product", "cost_note", "ads_note", "tacos_note", "stock_note", "returns_note",
"cogs_source", "fees", "sponsored_ads", "net_proceeds", "sales_note", "cogs_per_unit",
"cogs_total", "odoo_channel_sales", "odoo_channel_profit", "odoo_channel_margin",
"odoo_margin_source", "ad_impressions", "ad_clicks", "ad_spend", "ad_sales", "ad_orders",
"cvr", "cpc", "roas", "fba_on_hand", "inbound_units", "return_units", "sessions",
"unit_session_pct", "buy_box_pct", "listing_note", "units_30d", "net_sales_30d",
"ad_spend_30d", "health_note")
#: Columns a seeded view already written loses once; a person who shows one again keeps it.
_VIEW_COLUMNS_DROPPED = ("suggested_product", "suggested_units", "match_rule", "match_state")
def _column_before(columns, column):
index = list(columns).index(column)
return columns[index - 1] if index else None
def _insert_after(columns, before, column):
"""`columns` with `column` right after `before`, or appended when `before` is absent."""
out = [c for c in (columns or ()) if c != column]
out.insert(out.index(before) + 1 if before in out else len(out), column)
return out
# ---------------------------------------------------------------- refresh
def build_rows(con, period, row_links, thin=THIN_DEFAULT, cost_reader=None, today=None,
*, history=None, receipts=None, cost_problem=None, typed=None):
"""`(computed cells by SKU, report)` for one period. Every SKU in the mirror is present.
`typed` (`typed_costs`) fills COGS only for a SKU with no Odoo cost and no Odoo receipt (R43)."""
bounds = period_bounds(period, today)
econ = economics_by_sku(con, bounds["start"], bounds["end"])
names = titles(con)
codes = {link["code"] for link in row_links.values() if link.get("code")}
if cost_reader is None:
costs, problems = mirror_products(con, codes)
else:
costs, problems = cost_reader(codes) if codes else ({}, [])
allocations, usage = {}, []
if cost_reader is None:
if receipts is None and cost_problem is None:
try:
receipts = receipt_costs(con, [row["id"] for row in costs.values()]) if costs else []
except SourceUnavailable as exc:
cost_problem = str(exc)
if cost_problem:
problems.append(cost_problem)
elif costs:
allocations, usage = fifo_costs(con, bounds["start"], bounds["end"], row_links, costs,
receipts or [], (receipts or []) if history is None else history)
margins, margin_source = odoo_channel_margins(con, bounds["start"], bounds["end"])
history_by_product = {}
for observation in ((receipts or []) if history is None else history):
history_by_product.setdefault(observation["product_id"], []).append(observation)
history_texts = {pid: _history_text(observations, pid) for pid, observations in history_by_product.items()}
computed, linked, typed_used = {}, 0, 0
for item in econ:
item = dict(item, title=names.get(item["sku"]))
link = row_links.get(item["sku"]) or {}
found = costs.get(link.get("code") or "")
if found is not None:
linked += 1
row = compute_row(item, None if found is None or cost_reader is None else found["cost"],
link.get("units_per_sku") or 1, thin)
if found is not None:
channel = margins.get(found["id"])
row.update(odoo_channel_sales=channel["sales"] if channel else None,
odoo_channel_profit=channel["profit"] if channel else None,
odoo_channel_margin=channel["margin"] if channel else None,
odoo_margin_source=channel["source"] if channel else
("No Odoo channel sales" if margin_source == "Confirmed Fisch and Royal sales" else margin_source),
cost_history=history_texts.get(found["id"]))
if cost_reader is None:
allocation = allocations.get(item["sku"], {})
used = allocation.get("usage", [])
units = allocation.get("units", 0.0)
missing = allocation.get("missing", 0.0)
row["cost_source"] = ("Receipt cost unavailable" if found else
"Odoo product unavailable" if link.get("code") else "Odoo product not linked")
if allocation:
row["cost_receipt"] = ", ".join(dict.fromkeys(part["receipt"]["receipt"] for part in used)) or None
row["cost_coverage"] = (f"{units - missing:g} of {units:g} Odoo units | FIFO by sale date and SKU | "
f"Prior use: {allocation['prior_source']}")
if allocation.get("returns"):
row["cost_source"] = "Returns need receipt link"
elif allocation.get("incomplete_prior"):
row["cost_source"] = "Incomplete FIFO history"
elif missing > 1e-8:
row["cost_source"] = "Incomplete receipt coverage"
elif units > 0 and int(_num(item["units"])) > 0:
total = sum(part["purchase_used"] + part["landed_used"] for part in used)
row = {**row, **compute_row(item, total / int(item["units"]), 1, thin)}
sources = list(dict.fromkeys(part["receipt"]["source"] for part in used))
row["cost_source"] = ", ".join(sources)
else:
row = {**row, **compute_row(item, 0, 1, thin)}
row.update(cogs_per_unit=None, cost_source="No units sold")
typed_cost = (typed or {}).get(item["sku"])
receipt_known = found is not None and (cost_reader is not None or found["id"] in history_by_product)
use_typed = typed_cost is not None and row.get("cogs_total") is None and not receipt_known
if use_typed:
typed_used += 1
row = {**row, **compute_row(item, typed_cost, 1, thin), "cost_source": TYPED_COST_SOURCE}
if found is None:
row["odoo_margin_source"] = LINK_NOTE if not link.get("code") else PRODUCT_MISSING_NOTE
row["cost_note"] = cost_note(row, code=link.get("code"), found=found is not None or use_typed)
source = COGS_TYPED if use_typed else COGS_RECEIPT if row.get("cogs_total") is not None else None
computed[item["sku"]] = dict(cells(row, bounds["label"]), cogs_source=source)
report = {"period": bounds, "thin": thin, "skus": len(computed), "linked": linked,
"unlinked": len(computed) - linked, "typedCost": typed_used, "costProblems": problems,
"_costUsage": usage}
return computed, report
def refusal(checks, month):
"""One plain sentence naming the first failed validate() check, or None if all pass."""
label = dt.date.fromisoformat(f"{month}-01").strftime("%B %Y")
covered = (checks[0].get("covered") if checks else None) or {}
where = f"over {covered['label']}" if covered.get("partial") and covered.get("label") else f"in {label}"
sentences = (
lambda c: (f"Not verified: Amazon net proceeds miss settlements by "
f"{c.get('gap') or 0:.1%} {where}." if math.isfinite(c.get("gap") or 0)
else f"Not verified: Amazon has no settlements in {label} to tie to."),
lambda c: f"Not verified: the fee check cannot tell a wrong total from a right one in {label}.",
lambda c: f"Not verified: {c.get('a')} settlement rows in {label} have no readable date.",
lambda c: f"Not verified: SKU net proceeds do not add up to Amazon's total in {label}.",
lambda c: "Not verified: the SKU list is incomplete.",
)
for index, check in enumerate(checks):
if not check.get("ok"):
if index == 0 and check.get("reason"):
return check["reason"]
if index < len(sentences):
return sentences[index](check)
return f"Not verified: the Amazon figures do not tie in {label}."
return None
def _and(names):
return names[0] if len(names) == 1 else ", ".join(names[:-1]) + " and " + names[-1]
def withheld_parts(checks, month):
"""`{line, components, keys, month}` when only the settlement tie failed and sales or fees tied, else None.
Each failed component blanks its own figures and every profit figure; the tied figures publish."""
if not checks or checks[0].get("ok") or not all(c.get("ok") for c in checks[1:]):
return None
first = checks[0]
if "covered" in first and (first.get("covered") or {}).get("start") is None:
return None
parts = {p.get("component"): bool(p.get("ok")) for p in first.get("components") or ()}
if not (parts.get("product sales") or parts.get("order fees")):
return None
failed = list(dict.fromkeys([name for name, ok in parts.items() if not ok]
+ list(first.get("failedComponents") or ())))
keys = []
for part, figures in _WITHHELD_BY_PART.items(): # an absent sales or order fee part counts as failed
if not parts.get(part, part not in ("product sales", "order fees")) or part in failed:
keys.extend(figures)
keys = tuple(dict.fromkeys(keys + list(_WITHHELD_PROFIT)))
label = dt.date.fromisoformat(f"{month}-01").strftime("%B %Y")
line = (f"Not verified: {_and(failed)} in {label} {'does' if len(failed) == 1 else 'do'} not tie to "
"Amazon settlements." if failed else refusal(checks, month))
return {"line": line, "components": failed, "keys": keys, "month": month}
def _withhold(cells_by_key, withheld):
"""Rows with the figures `withheld` names blanked and explained by its line."""
if not withheld:
return cells_by_key
keys = withheld["keys"]
return {k: _explain(_blank(v, keys), keys, withheld["line"]) for k, v in cells_by_key.items()}
def refusal_line(sentence, covered):
"""The 409 line: the refusal, plus the day the rest settles when the settlement cycle gives one."""
clear = (covered or {}).get("clear")
if not sentence or not clear:
return sentence
return f"{sentence} The rest settles about {_day_label(dt.date.fromisoformat(clear))}."
def keep_figures(st, *, line, period, month, covered, by="", settling=None):
"""A refused refresh: the stored rows, period and verified range stay; only the refusal note is written.
Saving the refused period here would label the kept figures with a period they do not cover."""
stamp = dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds")
write_settings(st, lastRefusal={"line": line, "period": period, "month": month, "at": stamp,
"by": by or WRITER, "settling": settling})
settings = read_settings(st)
return {"tableKey": TABLE_KEY, "kept": True, "refusal": line, "refusedPeriod": period,
"refusedAt": stamp, "verifiedMonth": month, "verified": None, "verifiedRange": covered,
"period": period_bounds(settings["period"]), "skus": settings.get("skus"),
"linked": settings.get("linked"), "settling": settling}
def apply_owned(st, bands, *, today, owner=None, skus=None):
"""Open, reopen and stamp owned items in one store write. Returns the counts."""
import core.user_tables as ut
counts = {}
def _set(cur):
counts.clear()
counts.update(opened=0, reopened=0, stamped=0)
table = cur.get(TABLE_KEY)
if table is None:
return cur
for row in (table.get("rows") or {}).values():
sku = str(row.get("sku") or "").strip()
if skus is not None and sku not in skus:
continue
before = str(row.get("status") or "").strip()
changes = owned_changes(row, bands.get(sku), today=today, owner=owner)
if not changes:
continue
for key, value in changes.items():
if value is None:
row.pop(key, None)
else:
row[key] = value
if changes.get("status") == "Open":
counts["reopened" if before else "opened"] += 1
else:
counts["stamped"] += 1
return cur
ut.update_with_activity(TABLE_KEY, _set, st=st, by=WRITER, via=VIA, row_ids=None)
return dict(counts)
# ---------------------------------------------------------------- metrics, rollups, projects
_FIELD_BY_KEY = {f["key"]: f for f in COMPUTED_FIELDS + ASIN_FIELDS + PARENT_FIELDS + PROJECT_FIELDS}
def _decimals(field):
return (field.get("format") or {}).get("decimals")
def stored_cell(field, value):
"""A raw value as the stored cell string for its field type; None means blank."""
if value is None or value == "" or field is None:
return None if value in (None, "") else str(value)
ftype = field.get("type")
if ftype == "chart":
return value if isinstance(value, str) else json.dumps(value, separators=(",", ":"))
if isinstance(value, bool) or not isinstance(value, (int, float)):
return str(value)
if not math.isfinite(float(value)):
return None
if ftype == "pct":
return f"{value * 100:.2f}"
if ftype == "int":
places = _decimals(field)
return f"{float(value):.{places}f}" if places else str(int(round(value)))
if ftype == "currency" or isinstance(value, float):
return f"{float(value):.2f}"
return str(value)
def raw_value(field, cell):
"""The raw value a stored cell holds: numbers as numbers, pct as a fraction."""
if cell is None or cell == "":
return None
ftype = (field or {}).get("type")
if ftype == "chart":
try:
return json.loads(cell) if isinstance(cell, str) else cell
except ValueError:
return None
if ftype in ("pct", "int", "currency"):
try:
number = float(cell)
except (TypeError, ValueError):
return None
if not math.isfinite(number):
return None
if ftype == "pct":
return number / 100.0
if ftype == "int" and not _decimals(field or {}) and number.is_integer():
return int(number)
return number
return cell
def _raw_row(cells_by_key):
return {key: raw_value(_FIELD_BY_KEY.get(key), value) for key, value in cells_by_key.items()}
def _metric_problem_types():
own = getattr(M, "SourceUnavailable", None)
return (SourceUnavailable,) + ((own,) if isinstance(own, type) and issubclass(own, Exception) else ())
def merge_metrics(computed, metrics):
"""SKU cells with the metric and score cells added; `metrics` holds raw values by SKU."""
out = {}
for sku, values in computed.items():
raw = _raw_row(values)
raw.update({k: v for k, v in (metrics.get(sku) or {}).items() if k in METRIC_KEYS})
if not raw.get("asin") and (metrics.get(sku) or {}).get("asin"):
raw["asin"] = metrics[sku]["asin"]
raw.update(M.score_row(raw))
raw.update(M.blank_notes(raw))
merged = dict(values)
for key in METRIC_KEYS + ("asin",):
if key in raw:
merged[key] = stored_cell(_FIELD_BY_KEY.get(key), raw[key])
merged.setdefault("health_note", None)
out[sku] = merged
return out
def _blank(cells_by_key, keys, period=None):
out = {**cells_by_key, **{k: None for k in keys if k in cells_by_key or k in _FIELD_BY_KEY}}
if period is not None:
out["period"] = period
return out
def _explain(cells_by_key, keys, reason):
"""Cells with every reason column whose figures `keys` blanked set to `reason`."""
blanked = set(keys)
return {**cells_by_key, **{note: reason for note, figures in BLANK_REASONS.items()
if note in cells_by_key and blanked & set(figures)}}
def _grain_cost_note(flat, children):
"""Why a rollup row's COGS, profit or margin is blank, from its SKU rows."""
if all(flat.get(k) is not None for k in ("cogs_total", "net_profit", "margin")):
return ""
costless = [c for c in children if c.get("net_profit") is None]
if any(c.get("cost_note") == LINK_NOTE for c in costless):
return GRAIN_LINK_NOTE
if costless or flat.get("net_profit") is None:
return GRAIN_COST_NOTE
return M.NO_NET_SALES_NOTE
def _flatten(item):
flat = {}
for key, value in (item or {}).items():
if key in ("sums", "ratios") and isinstance(value, dict):
flat.update(value)
else:
flat[key] = value
return flat
def grain_rows(rows, key, *, thin=THIN_DEFAULT, period_label="", fields=()):
"""`{asin or parent: cells}` from `M.rollup`, which sums first and recomputes every ratio.
Band follows the rollup's margin and the period cell names the period."""
keys = [f["key"] for f in fields]
by_sku = {str(r.get("sku")): r for r in rows if r.get("sku") is not None}
out = {}
for item in M.rollup(rows, key=key):
flat = _flatten(item)
value = str(flat.get(key) or flat.get("key") or "").strip()
if not value:
continue
flat[key] = value
flat["band"] = band_for(flat.get("net_profit"), flat.get("margin"), thin)
children = [by_sku[s] for s in str(flat.get("skus") or "").split(", ") if s in by_sku]
flat["cost_note"] = _grain_cost_note(flat, children)
flat["period"] = period_label
out[value] = {k: stored_cell(_FIELD_BY_KEY.get(k), flat.get(k)) for k in keys
if k not in ("projects", "projects_label") + PDP_CELL_KEYS + _ASIN_SOURCE_KEYS}
return out
def sku_pictures(computed, refs):
"""SKU cells with `picture` set to the stored ref of the row's ASIN, None when it has none."""
return {sku: dict(values, picture=refs.get(str(values.get("asin") or "").strip()))
for sku, values in computed.items()}
def grain_pictures(by_grain, refs):
"""Set `picture` on every ASIN and parent cell after the rollup; returns the counts.
A parent uses its own stored picture, else the picture of its child ASIN with the most units
sold in the period, ties by ASIN ascending. No picture leaves the cell None."""
asins = by_grain.get("asin") or {}
best = {}
for asin, values in sorted(asins.items()):
values["picture"] = refs.get(asin)
parent = str(values.get("parent_asin") or "").strip()
units = _num(values.get("units_sold"))
if parent and values["picture"] and (parent not in best or units > best[parent][0]):
best[parent] = (units, values["picture"])
own = child = 0
for parent, values in (by_grain.get("parent_asin") or {}).items():
values["picture"] = refs.get(parent) or (best.get(parent) or (0, None))[1]
own += bool(refs.get(parent))
child += bool(values["picture"]) and not refs.get(parent)
return {"asin": sum(1 for values in asins.values() if values["picture"]),
"parent": own + child, "parentOwn": own, "parentChild": child}
def rollup_checks(computed, by_grain):
"""Each rollup table's units sold equal the SKU units sold, and it holds every distinct key."""
checks = []
for key, label, cells_by_key in (("asin", "ASIN", by_grain.get("asin") or {}),
("parent_asin", "parent ASIN", by_grain.get("parent_asin") or {})):
grouped = [values for values in computed.values() if str(values.get(key) or "").strip()]
skus = sum(int(_num(values.get("units_sold"))) for values in grouped)
grains = sum(int(_num(values.get("units_sold"))) for values in cells_by_key.values())
distinct = len({str(values.get(key)).strip() for values in grouped})
checks.append({"check": f"amazon_profit: {label} table units sold equal the SKU table units sold",
"ok": skus == grains, "a": grains, "b": skus, "gap": abs(grains - skus)})
checks.append({"check": f"amazon_profit: {label} table holds every distinct {label}, no cap",
"ok": len(cells_by_key) == distinct, "a": len(cells_by_key), "b": distinct,
"gap": abs(len(cells_by_key) - distinct)})
return checks
#: Grain figure families tied per row to an independent SQL aggregate of the SKUs' Amazon rows (rule 8):
#: (figures, independent source, figure words, what a failed tie withholds on that grain table).
_TIE_FAMILIES = (
(("units_sold", "net_sales", "fees", "sponsored_ads", "net_proceeds"), "Amazon economics", "sales figures",
_UNVERIFIED_KEYS + ("return_rate", "unit_session_pct", "listing_health")),
(("ad_spend",), "Amazon ad charges", "ad spend figures",
("ad_spend", "cpc", "acos", "roas", "tacos", "ads_health", "score", "priority", "ad_spend_30d")),
(("ad_impressions", "ad_clicks", "ad_sales", "ad_orders"), "the Amazon Ads reports", "ads figures",
M.ADS_FIGURES + ("ads_health", "score", "priority")),
(("fba_on_hand", "inbound_units"), "Amazon inventory planning", "stock figures",
("fba_on_hand", "inbound_units", "days_of_cover", "stock_health", "score", "priority")),
(("return_units",), "the Amazon returns reports", "returned units",
("return_units", "return_rate", "listing_health", "score", "priority")),
(("sessions",), "the Amazon traffic report", "sessions",
("sessions", "unit_session_pct", "buy_box_pct", "listing_health", "score", "priority")),
(("cogs_total", "net_profit"), "the SKU costs", "cost figures",
("cogs_total", "net_profit", "margin", "band", "profit_health", "score", "priority")),
)
_COUNT_FIGURES = frozenset(("units_sold", "ad_impressions", "ad_clicks", "ad_orders", "fba_on_hand",
"inbound_units", "return_units", "sessions"))
def _cent(value):
return None if value is None else float(f"{float(value):.2f}")
def _sub_cent(value):
"""A raw sum with a part below one cent: two float sums of it may round to cents one cent apart."""
return value is not None and abs(float(value) * 100 - round(float(value) * 100)) > 1e-6
def _sku_figures(con, start, end, skus):
"""`{figure: {sku: value or None}}` read straight from the mirror, rounded as a SKU cell stores it.
A family whose source table is absent maps to None."""
from modules import ranking_diagnostic, restock_planner
out = {}
cols = _require(con, ECONOMICS, _ECONOMICS_REQUIRED)
ads = " + ".join(f'COALESCE(TRY_CAST("{c}" AS DOUBLE), 0)' for c in ads_columns(cols)) or "0"
rows = con.execute(
"SELECT msku, SUM(sales_net_units_sold), SUM(sales_net_product_sales_amount), SUM(fees_total_amount), "
f"SUM({ads}) FROM {ECONOMICS} WHERE left(CAST(start_date AS VARCHAR), 10) BETWEEN ? AND ? "
"AND msku IN (SELECT unnest(?)) GROUP BY msku", [start, end, skus]).fetchall()
econ = {r[0]: [_num(v) for v in r[1:]] for r in rows}
zero = [0.0, 0.0, 0.0, 0.0]
out["units_sold"] = {s: int(econ.get(s, zero)[0]) for s in skus}
out["net_sales"] = {s: _cent(econ.get(s, zero)[1]) for s in skus}
out["fees"] = {s: _cent(econ.get(s, zero)[2]) for s in skus}
out["sponsored_ads"] = {s: _cent(econ.get(s, zero)[3]) for s in skus}
out["net_proceeds"] = {s: _cent(e[1] - e[2] - e[3]) for s, e in ((s, econ.get(s, zero)) for s in skus)}
loose = {figure: {s for s in econ if _sub_cent(econ[s][index])}
for index, figure in ((1, "net_sales"), (2, "fees"), (3, "sponsored_ads"))}
loose["net_proceeds"] = {s for s in econ if _sub_cent(econ[s][1] - econ[s][2] - econ[s][3])}
loose["ad_spend"] = loose["sponsored_ads"]
union = M._ads_parts(con)
adrows = {} if union is None else {r[0]: r[1:] for r in con.execute(
f"SELECT sku, SUM(impressions), SUM(clicks), SUM(cost), SUM(sales), SUM(orders) FROM ({union}) a "
"WHERE day BETWEEN CAST(? AS DATE) AND CAST(? AS DATE) AND sku IN (SELECT unnest(?)) GROUP BY sku",
[start, end, skus]).fetchall()}
for index, figure in ((0, "ad_impressions"), (1, "ad_clicks"), (3, "ad_sales"), (4, "ad_orders")):
out[figure] = None if union is None else {
s: (M._int(adrows[s][index]) if figure in _COUNT_FIGURES else _cent(M._f(adrows[s][index])))
if s in adrows else None for s in skus}
loose["ad_sales"] = {s for s in adrows if _sub_cent(M._f(adrows[s][3]))}
if ads_columns(cols):
out["ad_spend"] = {s: _cent(econ.get(s, zero)[3]) for s in skus}
else:
out["ad_spend"] = None if union is None else {s: _cent(M._f(adrows[s][2])) if s in adrows else None
for s in skus}
loose["ad_spend"] = {s for s in adrows if _sub_cent(M._f(adrows[s][2]))}
out["_loose"] = loose
plan_cols = _columns(con, restock_planner.PLANNING)
if {"sku", "available"} <= plan_cols:
inbound = 'TRY_CAST("inbound_quantity" AS DOUBLE)' if "inbound_quantity" in plan_cols else "NULL"
fnsku = "max(NULLIF(TRIM(CAST(\"fnsku\" AS VARCHAR)), ''))" if "fnsku" in plan_cols else "NULL"
plan = {r[0]: r[1:] for r in con.execute(
f'SELECT TRIM(CAST("sku" AS VARCHAR)), {fnsku}, SUM(TRY_CAST("available" AS DOUBLE)), SUM({inbound}) '
f'FROM {restock_planner.PLANNING} WHERE {restock_planner._newest(plan_cols, restock_planner.PLANNING)} '
'AND TRIM(CAST("sku" AS VARCHAR)) IN (SELECT unnest(?)) GROUP BY 1', [skus]).fetchall()}
out["_pool"] = {s: (plan[s][0] if s in plan else None) or s for s in skus}
out["fba_on_hand"] = {s: M._int(plan[s][1]) if s in plan else None for s in skus}
out["inbound_units"] = {s: M._int(plan[s][2]) if s in plan else None for s in skus}
else:
out["fba_on_hand"] = out["inbound_units"] = None
returns = M._returns_parts(con)
if returns is None:
out["return_units"] = None
else:
got = dict(con.execute(f"SELECT sku, SUM(units) FROM ({returns}) r WHERE day BETWEEN CAST(? AS DATE) "
"AND CAST(? AS DATE) AND sku IN (SELECT unnest(?)) GROUP BY sku",
[start, end, skus]).fetchall())
out["return_units"] = {s: M._int(got.get(s) or 0) for s in skus}
pricing = ranking_diagnostic._pricing(con)
out["_traffic"] = None if pricing is None else {a: M._int((v or {}).get("sessions")) for a, v in pricing.items()}
return out
def grain_ties(con, start, end, computed, by_grain):
"""Each ASIN and parent ASIN figure that holds a number ties per row to the sum of its SKUs' Amazon
rows, read again from the mirror in SQL (rule 8). A blank cell is withheld and carries its reason."""
skus = sorted(str(s) for s in computed)
source = _sku_figures(con, start, end, skus) if skus else {}
checks = []
for grain, label in (("asin", "ASIN"), ("parent_asin", "Parent ASIN")):
rows = by_grain.get(grain) or {}
members = {}
for sku in skus:
value = str((computed[sku] or {}).get(grain) or "").strip()
if value:
members.setdefault(value, []).append(sku)
for figures, words, what, _withheld in _TIE_FAMILIES:
for figure in figures:
compared = withheld = 0
mismatched, worst = [], 0.0
for key, cells_by_key in rows.items():
got = M._f(cells_by_key.get(figure))
if got is None:
withheld += 1
continue
compared += 1
want = _tie_expected(figure, members.get(key, []), computed, source)
gap = math.inf if want is None else abs(got - want)
# Half a cent for the row's own rounding, plus one cent per member SKU whose raw sum
# has a part below a cent: two float sums of it may round one cent apart.
loose = (source.get("_loose") or {}).get(figure) or set()
tolerance = (0.01 if figure in _COUNT_FIGURES else
0.0051 + 0.0101 * sum(1 for sku in members.get(key, []) if sku in loose))
if not gap <= tolerance:
mismatched.append(key)
worst = max(worst, gap)
ok = not mismatched
checks.append({"check": f"amazon_profit: {label} {_FIELD_BY_KEY[figure]['label'].lower()} tie to "
f"{words} per row", "ok": ok, "grain": grain, "figure": figure,
"compared": compared, "withheld": withheld, "mismatched": len(mismatched),
"keys": mismatched[:5], "rowsFailed": mismatched,
"gap": None if not compared else round(worst, 4),
"reason": None if ok else f"Not verified: {label} {what} do not tie to {words}."})
return checks
def _tie_expected(figure, members, computed, source):
"""The independent value of one grain figure from its member SKUs, or None when no source holds it."""
if figure in ("cogs_total", "net_profit"):
values = [M._f((computed[s] or {}).get(figure)) for s in members]
active = [s for s in members if M._active(_raw_row(computed[s] or {}))]
if any(M._f((computed[s] or {}).get(figure)) is None for s in active):
return None
found = [v for v in values if v is not None]
return _cent(sum(found)) if found else None
if figure == "sessions":
traffic = source.get("_traffic")
if traffic is None:
return None
asins = dict.fromkeys(str((computed[s] or {}).get("asin") or "").strip() for s in members)
found = [traffic.get(a) for a in asins if a and traffic.get(a) is not None]
return float(sum(found)) if found else None
per_sku = source.get(figure)
if per_sku is None:
return None
if figure in ("fba_on_hand", "inbound_units"):
pools = {}
for sku in sorted(members):
pools.setdefault(source["_pool"].get(sku) or sku, per_sku.get(sku))
found = [v for v in pools.values() if v is not None]
else:
found = [per_sku.get(s) for s in members if per_sku.get(s) is not None]
return float(sum(found)) if found else None
def _withhold_untied(by_grain, checks):
"""On each row whose tie failed, blank what that tie withholds and write the failed sentence as the reason."""
withheld = {figure: keys for figures, _w, _what, keys in _TIE_FAMILIES for figure in figures}
for check in checks:
if check.get("ok") or check.get("grain") not in by_grain:
continue
keys, cells = withheld[check["figure"]], by_grain[check["grain"]]
for row in check.get("rowsFailed") or ():
if row in cells:
cells[row] = _explain(_blank(cells[row], keys), keys, check["reason"])
return by_grain
def _grain_period_keys(fields, key_column):
return tuple(f["key"] for f in fields
if f["key"] not in (key_column, "picture", "parent_asin", "title", "projects", "projects_label", "history")
+ COMPETITOR_KEYS + PDP_CELL_KEYS + _ASIN_SOURCE_KEYS)
def _listing_dates(con):
"""Read the newest Amazon lastUpdatedDate per ASIN from raw listing summaries."""
if "summaries" not in _columns(con, "amzn_listing_items"):
return {}
rows = con.execute("""SELECT json_extract_string(s.value, '$.asin'),
CAST(MAX(TRY_CAST(json_extract_string(s.value, '$.lastUpdatedDate') AS TIMESTAMPTZ))
AT TIME ZONE 'UTC' AS DATE)
FROM amzn_listing_items l, json_each(TRY_CAST(l.summaries AS JSON)) s GROUP BY 1""").fetchall()
return {asin: date.isoformat() for asin, date in rows if asin and date is not None}
def _exposure_detail(exposure):
"""Format the recorded exposure multiple with the prototype's one-decimal rounding."""
if not exposure or exposure["multiple"] is None:
return None
multiple = Decimal.from_float(float(exposure["multiple"])).quantize(Decimal("0.1"), rounding=ROUND_HALF_UP)
return f"{multiple}x {exposure['base']} share"
def _asin_signal_cells(signal, history):
"""Store scalar cells and raw monthly bars for every reported historical month."""
out = {key: stored_cell(_FIELD_BY_KEY[key], signal[key]) for key in ("trend_score", "funnel_score")}
out.update({key + "_reason": value for key, value in signal["reasons"].items()
if key in ("trend_score", "funnel_score", "exposure")})
exposure = signal["exposure"]
out["exposure"] = EXPOSURES[exposure["kind"]] if exposure else None
out["exposure_label"] = _exposure_detail(exposure)
for kind, key, _label, _now_type in SEARCH_TRENDS:
trend = signal["trends"][kind]
points = {point["month"]: point for point in trend["series"]}
series = [points[month] for month in history]
latest = series[-1] if series else {"value": None, "reason": signal["note"] or "No search report"}
now = latest["value"] if kind == "search" else signal[kind]
out.update({key: None if trend["change"] is None else str(trend["change"] * 100),
key + "_series": json.dumps(series, separators=(",", ":")),
key + "_now": (stored_cell(_FIELD_BY_KEY[key + "_now"], now) if kind == "search"
else None if now is None else str(now * 100)),
key + "_reason": trend["reason"],
key + "_now_reason": (signal["note"] or latest["reason"] if kind == "search"
else signal["reasons"][kind])})
return out
def paint_asin_signals(st, con, *, today):
"""Refresh source signals and Amazon's date for every stored ASIN, independent of the profit period."""
from modules import amazon_search_signals as signals
asins = [str(row["asin"]).strip() for row in table_rows(st, ASIN_KEY).values() if row.get("asin")]
result = signals.signals(con, asins, today=today)
dates = _listing_dates(con)
parts = _part_dates.cells(st, asins)
computed = {asin: dict(_asin_signal_cells(signal, result["history"]), **parts[asin], pdp_last_updated=dates.get(asin),
pdp_last_updated_reason=None if asin in dates else "No listing lastUpdatedDate reported")
for asin, signal in result["asins"].items()}
return write_rows(st, computed, table_key=ASIN_KEY, key_column="asin", label=ASIN_LABEL, period_keys=())
def paint_asin_part_dates(st):
"""Refresh recorded part dates without profit validation, search reports or Odoo matching."""
asins = [str(row['asin']).strip() for row in table_rows(st, ASIN_KEY).values() if row.get('asin')]
if not asins:
return {'rows': 0}
return write_rows(st, _part_dates.cells(st, asins), table_key=ASIN_KEY,
key_column='asin', label=ASIN_LABEL, period_keys=())
def asin_signal_checks(st, con, *, today=None):
"""Tie stored signal cells to the signal reader's independently validated raw monthly inputs."""
from modules import amazon_search_signals as signals
rows = {str(row["asin"]).strip(): row for row in table_rows(st, ASIN_KEY).values() if row.get("asin")}
result = signals.signals(con, rows, today=today)
verified = signals.validate(con, result, today=today)
differences = sum(any(rows[asin].get(key) != value for key, value in _asin_signal_cells(signal, result["history"]).items())
for asin, signal in result["asins"].items())
return [{"check": "amazon_profit: stored ASIN search signals tie to raw monthly reports",
"ok": verified["ok"] and not differences, "rows": len(rows), "differences": differences,
"reason": verified["reason"]}]
def listing_date_checks(st, con):
"""Independently tie stored PDP dates to the newest raw summary timestamp per ASIN."""
newest = {}
available = "summaries" in _columns(con, "amzn_listing_items")
if available:
cursor = con.execute("SELECT summaries FROM amzn_listing_items")
for batch in iter(lambda: cursor.fetchmany(2048), []):
for (raw,) in batch:
try:
summaries = json.loads(raw) if isinstance(raw, str) else raw
except (TypeError, ValueError):
continue
for summary in summaries if isinstance(summaries, list) else ():
if not isinstance(summary, dict) or not summary.get("asin"):
continue
try:
stamp = dt.datetime.fromisoformat(str(summary.get("lastUpdatedDate")).replace("Z", "+00:00"))
except ValueError:
continue
stamp = stamp.replace(tzinfo=stamp.tzinfo or dt.timezone.utc).astimezone(dt.timezone.utc)
asin = summary["asin"]
if asin not in newest or stamp > newest[asin]:
newest[asin] = stamp
rows = list(table_rows(st, ASIN_KEY).values())
differences = sum(row.get("pdp_last_updated") != (newest[row["asin"]].date().isoformat()
if row.get("asin") in newest else None) for row in rows)
return [{"check": "amazon_profit: PDP last updated ties to newest listing lastUpdatedDate per ASIN",
"ok": available and not differences, "rows": len(rows), "differences": differences,
"reason": None if available else "Listing summaries are unavailable"}]
def _date(value):
try:
return dt.date.fromisoformat(str(value or "").strip()[:10])
except ValueError:
return None
def project_health(row, today):
"""Done, Overdue, Due this week or On track, from status and the deadline."""
if str(row.get("status") or "").strip() == "Done":
return "Done"
deadline = _date(row.get("deadline_on"))
if deadline is None:
return "On track"
if deadline < today:
return "Overdue"
return "Due this week" if (deadline - today).days <= DUE_SOON_DAYS else "On track"
def _figure(result, metric):
found = next((f for f in result.get("figures") or () if f.get("id") == metric), None) or {}
return (found.get("current") or {}).get("value"), (found.get("comparison") or {}).get("value"), found
#: `impact_state` reaches a screen, so a results token reads as words. Unknown tokens pass through.
IMPACT_STATE_WORDS = {
"missing_grain_or_key": "Needs a grain and a key",
"not_verified": "Not verified: the source totals did not tie",
"unavailable": "Source unavailable. Sync Amazon.",
"waiting_for_completion": "Completes first",
"waiting_for_completed_day": "Waits for the first full day after completion",
"no_prior_data": "No sales one year earlier to compare against",
"partial_data": "Only part of the window has data",
}
def project_cells(rid, row, con, today):
"""Computed cells for one project; impact only for a Done project with a completion date."""
from modules import sku_action_results as SR
out = {k: None for k in PROJECT_COMPUTED_KEYS}
out["health"] = project_health(row, today)
completed = str(row.get("completed_on") or "").strip()[:10]
if str(row.get("status") or "").strip() != "Done" or not completed:
return out
grain, key = M.grain_key(row.get("grain")), str(row.get("key") or "").strip()
if not grain or not key:
out["impact_state"] = IMPACT_STATE_WORDS["missing_grain_or_key"]
return out
action = {"id": str(rid), "grain": grain, "key": key, "status": "done", "completed_on": completed}
try:
result = SR.results(con, action, today=today)
except SR.ValidationFailed:
out["impact_state"] = IMPACT_STATE_WORDS["not_verified"]
return out
except (SR.SourceUnavailable, ValueError):
out["impact_state"] = IMPACT_STATE_WORDS["unavailable"]
return out
except Exception as exc: # noqa: BLE001
# One project that cannot be measured must not stop the refresh of every table.
_log(f"project {rid}: impact unavailable ({type(exc).__name__}: {exc})")
out["impact_state"] = IMPACT_STATE_WORDS["unavailable"]
return out
sales_now, sales_prior, sales = _figure(result, "sales")
units_now, units_prior, _units = _figure(result, "units")
roas_now, roas_prior, _roas = _figure(result, "roas")
raw = {"net_sales_since": sales_now, "net_sales_prior": sales_prior, "units_since": units_now,
"units_prior": units_prior, "roas_since": roas_now, "roas_prior": roas_prior,
"impact": None if sales_now is None or sales_prior is None else sales_now - sales_prior,
"impact_units": None if units_now is None or units_prior is None else units_now - units_prior}
out.update({k: stored_cell(_FIELD_BY_KEY[k], v) for k, v in raw.items()})
if out["impact"] is None:
state = result.get("state") or "unavailable"
if state == "ready":
state = (sales.get("comparison") or {}).get("status") or (sales.get("current") or {}).get("status") \
or "unavailable"
out["impact_state"] = state
out["impact_state"] = IMPACT_STATE_WORDS.get(out["impact_state"], out["impact_state"])
return out
def migrate_projects(st):
"""Copy each legacy project row once into the projects store."""
from core import amazon_projects
rows = table_rows(st, PROJECTS_KEY)
return {"rows": len(rows), "copied": amazon_projects.move_rows(st, rows)}
def paint_project_cells(st, con):
"""Repaint project progress after refresh adds rows or updates their lineage."""
from core import amazon_projects as AP
import core.user_tables as ut
tables = ((TABLE_KEY, ("sku", "asin", "parent_asin")), (ASIN_KEY, ("asin", "parent_asin")),
(PARENT_KEY, ("parent_asin",)))
source = {grains[0]: table_rows(st, key) for key, grains in tables}
lineage = M.lineage_rows(con, source)
planned = []
for key, grains in tables:
rows = source[grains[0]]
chains = {rid: [AP.target(g, str(row[g]).strip()) for g in grains if str(row.get(g) or "").strip()]
for rid, row in lineage[grains[0]].items()}
planned.append((key, rows, chains))
pairs = sorted({pair for _key, _rows, chains in planned for chain in chains.values() for pair in chain})
found = AP.projects_for(st, pairs) if pairs else {}
counts = {}
for key, rows, chains in planned:
wanted = {}
for rid, chain in chains.items():
records = {p["id"]: p for pair in chain for p in found.get(pair, ())}
counted = AP.summary(records.values())
values = {"projects": stored_cell({"type": "pct"}, counted["average"] / 100)
if counted["total"] else None,
"projects_label": f"{counted['completed']} of {counted['total']} completed"
if counted["total"] else None}
if any(rows[rid].get(k) != value for k, value in values.items()):
wanted[rid] = values
if not wanted:
continue
def _set(cur, key=key, wanted=wanted):
saved = ((cur or {}).get(key) or {}).get("rows") or {}
for rid, values in wanted.items():
if rid in saved:
_apply_cells(saved[rid], values)
return cur
ut.update_with_activity(key, _set, st=st, by=WRITER, via=VIA, row_ids=list(wanted))
counts[key] = len(wanted)
return counts
def _evaluate_project_filter_tasks(st, *, today, by):
"""Evaluate saved task filters with their authors' permissions after project publication."""
from importlib import import_module
try:
tasks = import_module("core.amazon_project_tasks")
filters = import_module("modules.amazon_project_filters")
except ModuleNotFoundError as exc:
if exc.name not in ("core.amazon_project_tasks", "modules.amazon_project_filters"):
raise
_log(f"Project filter tasks need {exc.name}. Combine prerequisite: merge lane D's task evaluator and matcher.")
return None
return tasks.evaluate_filter_tasks(st, filters.matching_keys, today=today, by=by)
def update_projects(st, con, *, today, computed, period, period_label, filter_task_evaluator=None):
"""Publish independently verified project period totals and Impact."""
from core import amazon_projects as AP
from modules import sku_action_results as SR
keys = ("net_sales", "net_profit", "units_sold")
supplied = {}
for sku, row in computed.items():
stored = cells({key: row.get(key) for key in keys}, period_label)
supplied[sku] = dict(row, **{key: stored[key] for key in keys})
def impact_reader(action, *, today):
try:
return SR.results(con, action, today=today)
except SR.ValidationFailed:
return {"state": "not_verified"}
except (SR.SourceUnavailable, ValueError):
return {"state": "unavailable"}
result = AP.update_projects(st, supplied, period=period, period_label=period_label,
today=today, impact_reader=impact_reader,
impact_state_words=IMPACT_STATE_WORDS)
evaluator = _evaluate_project_filter_tasks if filter_task_evaluator is None else filter_task_evaluator
evaluator(st, today=today, by=WRITER)
return result
def write_competitor_cells(st, plan):
"""Write only the quote-owned cells and rows in a validated plan."""
import core.user_tables as ut
for grain, table_key in (("sku", TABLE_KEY), ("asin", ASIN_KEY), ("parent_asin", PARENT_KEY)):
wanted = plan.get(grain) or {}
if not wanted:
continue
def _set(cur, table_key=table_key, wanted=wanted):
rows = ((cur or {}).get(table_key) or {}).get("rows") or {}
for rid, values in wanted.items():
if rid in rows:
_apply_cells(rows[rid], values)
return cur
ut.update_with_activity(table_key, _set, st=st, by=WRITER, via=VIA, row_ids=list(wanted))
def on_quotes(tenant_key, changed_asins):
"""Validate and repaint the competitor cells affected by published quotes."""
from harness import amazon_store
try:
st = amazon_store._runtime_for(tenant_key)
if st is None or not changed_asins:
return {"state": "skipped"}
tables = {"sku": table_rows(st, TABLE_KEY), "asin": table_rows(st, ASIN_KEY),
"parent_asin": table_rows(st, PARENT_KEY)}
from harness import runtime as R
con = R.mirror_cursor(st, require_ready=False)
try:
lineage = M.lineage_rows(con, tables, children=True)
finally:
con.close()
plan = _rivals.cells(st, lineage, changed_asins)
# A moved child must also clear its former parent's persisted rollup.
hit = _rivals.affected(st, changed_asins)
old = _rivals.family(tables["asin"])
stale = {rid for rid, row in tables["parent_asin"].items()
if hit.intersection(old.get(_rivals._maybe_asin(row.get("parent_asin")), ()))} - set(plan["parent_asin"])
if stale:
full = _rivals.cells(st, lineage)
plan["parent_asin"].update({rid: full["parent_asin"][rid] for rid in stale})
plan["asin"] = {rid: values for rid, values in plan["asin"].items() if rid in tables["asin"]}
checks = _rivals.validate(st, lineage, plan)
failed = [c for c in checks if not c.get("ok")]
if failed:
_log("competitor cells not verified: " + ", ".join(str(c.get("check")) for c in failed))
return {"state": "not_verified"}
write_competitor_cells(st, plan)
return {grain: len(rows) for grain, rows in plan.items()}
except Exception as exc:
_log(f"competitor cells unavailable ({type(exc).__name__})")
return {"state": "failed"}
def refresh(st, con, *, period=None, thin=None, by="", cost_reader=None, today=None,
tenant=TENANT_ZERO, validator=None):
"""Recompute every SKU for the period and store it; owned cells stay. Returns the report.
It runs validate() first for `verification_month`, stepped back by `settled_month` while it settles.
A failed settlement component publishes the tied figures and withholds the rest (`withheld_parts`).
Any other failed tie writes no rows and no period (`keep_figures`): the report carries `refusal`."""
settings = read_settings(st)
period = str(period or settings["period"]).strip().lower()
period_bounds(period, today)
thin = settings["thin"] if thin is None else clean_thin(thin)
day = today or _today()
month, settling = verification_month(period, day), None
try:
month, settling = settled_month(con, month, economics_currency(con))
except SourceUnavailable:
pass # validate() names the missing source
checks = (validator or (lambda c, m: validate(month=m, con=c)))(con, month)
sentence = refusal(checks, month)
covered = next((c["covered"] for c in checks if isinstance(c.get("covered"), dict)), None)
partial = withheld_parts(checks, month)
if sentence and partial is None: # the settlement check runs before every write
with _LOCK:
return keep_figures(st, line=refusal_line(sentence, covered), period=period, month=month,
covered=covered, by=by, settling=settling)
verified = covered.get("sentence") if covered else None
with _LOCK:
ensure_table(st, by)
merged = merge_duplicates(st)
decisions = apply_decisions(st, by=by or WRITER)
owner, owner_label = default_owner(st, tenant)
for spec in table_specs():
ensure_views(st, owner or by, table_key=spec["key"], template_key=spec["template"])
row_links = links(table_rows(st))
receipts, history, cost_problem = None, [], None
if cost_reader is None:
history = read_cost_history(st)
products, _problems = mirror_products(con, {link["code"] for link in row_links.values() if link["code"]})
if products:
try:
receipts = receipt_costs(con, [row["id"] for row in products.values()])
except SourceUnavailable as exc:
cost_problem = str(exc)
else:
history = append_cost_history(st, receipts)
typed = typed_costs(table_rows(st))
computed, report = build_rows(con, period, row_links, thin, cost_reader, today,
history=history, receipts=receipts, cost_problem=cost_problem, typed=typed)
usage = report.pop("_costUsage")
if cost_reader is None:
landed_checks = landed_cost_checks(con, usage, history)
report["landedValidation"] = landed_checks
if not all(check["ok"] for check in landed_checks):
return dict(keep_figures(st, line="Not verified: receipt costs do not tie to Odoo landed cost lines.",
period=period, month=month, covered=covered, by=by),
landedValidation=landed_checks)
typed_checks = typed_cogs_checks(con, computed, typed, usage, report["period"])
report["typedValidation"] = typed_checks
if not all(check["ok"] for check in typed_checks):
return dict(keep_figures(st, line=TYPED_REFUSAL, period=period, month=month, covered=covered, by=by),
typedValidation=typed_checks)
bounds = report["period"]
metrics, metric_problem = {}, None
try:
metrics = M.sku_metrics(con, start=bounds["start"], end=bounds["end"], skus=list(computed), today=day)
except _metric_problem_types() as exc:
metric_problem = str(exc)
computed = _withhold(merge_metrics(computed, metrics), partial)
metric_checks = [] if metric_problem else list(
M.validate_metrics(con, start=bounds["start"], end=bounds["end"], computed=metrics) or [])
metric_refusal = metric_problem or next(
(f"Not verified: {c.get('check')}." for c in metric_checks if not c.get("ok")), None)
if metric_refusal:
computed = {sku: _explain(_blank(values, _METRIC_VALUE_KEYS), _METRIC_VALUE_KEYS, metric_refusal)
for sku, values in computed.items()}
from harness import amazon_pictures
refs = amazon_pictures.picture_refs(st) or {}
computed = sku_pictures(computed, refs)
report.update(write_rows(st, computed), merged=merged,
decisions={k: decisions[k] for k in ("confirmed", "rejected")})
report["projects"] = update_projects(st, con, today=day, computed=computed,
period=period, period_label=bounds["label"])
report["projectValidation"] = report["projects"]["validation"]
rows = [dict(_raw_row(values), sku=sku, velocity_30d=(metrics.get(sku) or {}).get("velocity_30d"),
fnsku=(metrics.get(sku) or {}).get("fnsku"))
for sku, values in computed.items()]
by_grain = {key: grain_rows(rows, key, thin=thin, period_label=bounds["label"], fields=fields)
for key, fields in (("asin", ASIN_FIELDS), ("parent_asin", PARENT_FIELDS))}
report["pictures"] = dict(grain_pictures(by_grain, refs), stored=len(refs),
sku=sum(1 for values in computed.values() if values.get("picture")))
for key, cells_by_key in list(by_grain.items()):
cells_by_key = _withhold(cells_by_key, partial)
if metric_refusal:
cells_by_key = {k: _explain(_blank(v, _METRIC_VALUE_KEYS), _METRIC_VALUE_KEYS, metric_refusal)
for k, v in cells_by_key.items()}
by_grain[key] = cells_by_key
tie = rollup_checks(computed, by_grain)
if not all(check["ok"] for check in tie):
failed = "Not verified: the rollup totals do not add up to the SKU totals."
by_grain = {key: {k: _explain(_blank(v, _UNVERIFIED_KEYS, failed), _UNVERIFIED_KEYS, failed)
for k, v in cells_by_key.items()}
for key, cells_by_key in by_grain.items()}
ties = grain_ties(con, bounds["start"], bounds["end"], computed, by_grain)
by_grain = _withhold_untied(by_grain, ties)
report["metricValidation"] = metric_checks + tie + ties
report["metricRefusal"] = metric_refusal
report["asin"] = write_rows(st, by_grain["asin"], table_key=ASIN_KEY, key_column="asin",
label=ASIN_LABEL, period_keys=_grain_period_keys(ASIN_FIELDS, "asin"))
report["parent"] = write_rows(st, by_grain["parent_asin"], table_key=PARENT_KEY,
key_column="parent_asin", label=PARENT_LABEL,
period_keys=_grain_period_keys(PARENT_FIELDS, "parent_asin"))
report["asinSignals"] = paint_asin_signals(st, con, today=day)
report["blankReasons"] = blank_check(stored_tables(st))
bands = {sku: values.get("band") for sku, values in computed.items()}
report["owned"] = apply_owned(st, bands, today=day, owner=owner_label)
stamp = dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds")
write_settings(st, period=period, thin=thin, refreshedAt=stamp, refreshedBy=by or WRITER,
lastPeriod=report["period"], skus=report["skus"], linked=report["linked"],
verifiedMonth=month, refusal=None, verified=verified, verifiedRange=covered,
lastRefusal=None, withheld=partial, settling=settling)
report.update(tableKey=TABLE_KEY, refreshedAt=stamp, verifiedMonth=month, refusal=None,
verified=verified, verifiedRange=covered, defaultOwner=owner,
withheld=partial, settling=settling)
try:
report["projectCells"] = paint_project_cells(st, con)
except Exception as exc:
_log(f"project cells unavailable ({type(exc).__name__})")
report["dataEntry"] = data_entry_sync(st, con, by=by)
return report
def data_entry_sync(st, con, *, by=""):
"""Rebuild Data Entry after a refresh. A failure there is reported and never fails the refresh."""
from modules import amazon_data_entry
try:
return amazon_data_entry.sync(st, con, by=by)
except Exception as exc: # noqa: BLE001
return {"error": f"Data Entry was not rebuilt: {exc}"}
def _stored_day(settings):
"""The `today` the stored figures were computed on, so a rolling window covers the same days."""
end = (settings.get("lastPeriod") or {}).get("end")
try:
return dt.date.fromisoformat(end) + dt.timedelta(days=1) if end else None
except (TypeError, ValueError):
return None
def recompute_sku(st, con, sku, *, by="", cost_reader=None, today=None):
"""Recompute one SKU's cost and profit from its current Odoo link and write that row only.
It keeps the stored period and window, so it needs no new settlement tie and a refused
refresh never blocks it. The link door, the Link status cell hook and Data Entry call it.
Returns `{sku, written, row}`, or `{sku, refusal}` when receipt costs do not tie."""
import core.user_tables as ut
sku = str(sku or "").strip()
with _LOCK:
settings = read_settings(st)
day = today or _stored_day(settings) or _today()
stored = {str(r.get("sku") or "").strip(): r for r in table_rows(st).values()}
if not sku or sku not in stored:
return {"sku": sku, "written": False, "row": None}
row_links = links(table_rows(st))
receipts, history, cost_problem = None, [], None
if cost_reader is None:
history = read_cost_history(st)
products, _problems = mirror_products(con, {link["code"] for link in row_links.values() if link["code"]})
if products:
try:
receipts = receipt_costs(con, [row["id"] for row in products.values()])
except SourceUnavailable as exc:
cost_problem = str(exc)
else:
history = append_cost_history(st, receipts)
# Every link goes in: FIFO shares receipts across SKUs on one Odoo product.
typed = typed_costs(table_rows(st))
computed, report = build_rows(con, settings["period"], row_links, settings["thin"], cost_reader, day,
history=history, receipts=receipts, cost_problem=cost_problem, typed=typed)
usage = report.pop("_costUsage")
if cost_reader is None and not all(c["ok"] for c in landed_cost_checks(con, usage, history)):
return {"sku": sku, "refusal": "Not verified: receipt costs do not tie to Odoo landed cost lines."}
if not all(c["ok"] for c in typed_cogs_checks(con, computed, typed, usage, report["period"])):
return {"sku": sku, "refusal": TYPED_REFUSAL}
values = computed.get(sku)
if values is None:
return {"sku": sku, "written": False, "row": dict(stored[sku])}
before = stored[sku]
if any(before.get(k) not in (None, "") for k in _METRIC_VALUE_KEYS):
# The stored metric inputs passed their checks at the last refresh; only the score moves.
inputs = {k: raw_value(_FIELD_BY_KEY.get(k), before.get(k)) for k in METRIC_KEYS if k in before}
values = merge_metrics({sku: values}, {sku: inputs})[sku]
values = {k: v for k, v in values.items() if k not in _IDENTITY_KEYS or k == "title"}
values = _withhold({sku: values}, settings.get("withheld"))[sku]
after = {}
def _set(cur):
after.clear()
table = (cur or {}).get(TABLE_KEY)
if table is None:
return cur
for row in (table.get("rows") or {}).values():
if str(row.get("sku") or "").strip() == sku:
_apply_cells(row, values)
after.update(row)
return cur
ut.update_with_activity(TABLE_KEY, _set, st=st, by=by or WRITER, via=VIA, row_ids=None)
_recompute_stored_grains(st, after, settings, con)
_owner, owner_label = default_owner(st, getattr(st, "key", TENANT_ZERO))
apply_owned(st, {sku: after.get("band")}, today=today or _today(), owner=owner_label, skus={sku})
after = next((row for row in table_rows(st).values() if row.get("sku") == sku), after)
write_settings(st, linked=report["linked"])
return {"sku": sku, "written": True, "row": dict(after)}
def _recompute_stored_grains(st, changed, settings, con=None):
"""Update the edited SKU's rollups from persisted cells after their sums tie.
With `con`, each rewritten row also ties to the mirror; a failed figure stays withheld with its reason."""
period_label = str(changed.get("period") or "")
if not period_label:
return
source = {rid: row for rid, row in table_rows(st).items() if row.get("period") == period_label}
rows = [_raw_row(row) for row in source.values()]
plans = {grain: _withhold(grain_rows(rows, grain, thin=settings["thin"], period_label=period_label,
fields=fields), settings.get("withheld"))
for grain, fields in (("asin", ASIN_FIELDS), ("parent_asin", PARENT_FIELDS))}
checks = rollup_checks({rid: row for rid, row in source.items()}, plans)
if not all(check["ok"] for check in checks):
_log("edited SKU rollups not verified")
return
bounds = settings.get("lastPeriod") or {}
if con is not None and bounds.get("start") and bounds.get("end"):
plans = {grain: {key: cells for key, cells in plans[grain].items()
if key == str(changed.get(grain) or "").strip()} for grain in plans}
sku_cells = {str(r.get("sku")).strip(): r for r in source.values() if str(r.get("sku") or "").strip()}
plans = _withhold_untied(plans, grain_ties(con, bounds["start"], bounds["end"], sku_cells, plans))
for grain, table_key, fields in (("asin", ASIN_KEY, ASIN_FIELDS), ("parent_asin", PARENT_KEY, PARENT_FIELDS)):
key = str(changed.get(grain) or "").strip()
if key and key in plans[grain]:
values = {k: v for k, v in plans[grain][key].items() if k not in _IDENTITY_KEYS or k == grain}
write_rows(st, {key: values}, table_key=table_key, key_column=grain, period_keys=())
# ---------------------------------------------------------------- validation source
_SNAPSHOT = {}
def _mirror_has_economics(con):
try:
return bool(_columns(con, ECONOMICS))
except Exception: # noqa: BLE001
return False
def snapshot_cursor(tenant=TENANT_ZERO, tables=(ECONOMICS, SETTLEMENTS, LISTINGS, "amzn_orders",
"amzn_listing_items", "amzn_search_query_performance",
"amzn_ads_sp_advertised_products", "amzn_ads_sb_advertised_products",
"amzn_ads_sd_advertised_products")):
"""An in-memory DuckDB over the stored mirror snapshot. Downloads only; writes nothing back."""
key = (tenant, tuple(tables))
if key in _SNAPSHOT:
return _SNAPSHOT[key].cursor()
import duckdb
from harness import runtime as R
from harness import mirror_store
rt = R.get_runtime(tenant)
mirror = mirror_store.for_connector("amazon")
raw = mirror._download(rt, "manifest.json")
if raw is None:
raise SourceUnavailable(f"{tenant} has no stored Amazon mirror.")
manifest = json.loads(raw)
if manifest.get("tenant") != tenant:
raise SourceUnavailable("The stored Amazon mirror names another tenant.")
folder = Path(tempfile.mkdtemp(prefix="amazon-profit-"))
atexit.register(shutil.rmtree, folder, True)
con = duckdb.connect()
for name in tables:
meta = (manifest.get("tables") or {}).get(name)
if not isinstance(meta, dict):
continue
paths = []
for index, part in enumerate(meta.get("parts") or [meta]):
blob_key = part.get("key") or f"{name}.parquet"
data = mirror._download(rt, blob_key)
if data is None or (part.get("sha256") and hashlib.sha256(data).hexdigest() != part["sha256"]):
raise SourceUnavailable(f"The stored {name} snapshot failed its checksum.")
path = folder / f"{name}.{index:04d}.parquet"
path.write_bytes(data)
paths.append(str(path))
if paths:
listed = ", ".join("'" + p.replace("'", "''") + "'" for p in paths)
con.execute(f"CREATE VIEW {name} AS SELECT * FROM read_parquet([{listed}], union_by_name=true)")
_SNAPSHOT[key] = con
return con.cursor()
def validation_cursor(tenant=TENANT_ZERO):
"""`(cursor, source)`: the local mirror when it holds economics, else the stored snapshot."""
try:
from harness import datastore as DS
if Path(DS.DB_PATH).is_file():
con = DS.ro_cursor()
if _mirror_has_economics(con):
return con, "local mirror"
except Exception: # noqa: BLE001
pass
return snapshot_cursor(tenant), "stored snapshot"
# ---------------------------------------------------------------- validate
def _tie(a, b, tolerance=SETTLEMENT_TOLERANCE):
gap = abs(a - b) / abs(b) if b else (0.0 if a == b else math.inf)
return gap <= tolerance, gap
def landed_cost_checks(con, usage=None, history=None):
"""Tie allocated receipt costs to independent SQL sums of valuations and posted adjustment lines."""
if not usage:
return [{"check": "amazon_profit: no receipt COGS used", "ok": True,
"state": "unused", "verified": False, "a": 0, "b": 0, "gap": 0}]
try:
_require_odoo(con, _RECEIPT_COLUMNS)
except SourceUnavailable as exc:
return [{"check": "amazon_profit: receipt cost source unavailable", "ok": False,
"state": "unavailable", "reason": str(exc), "a": None, "b": None, "gap": None}]
saved = {row["observation_id"]: row for row in history or []}
base = {(int(pid), int(move)): (purchase, float(quantity), name, day)
for pid, move, purchase, quantity, name, day in con.execute(
f"SELECT v.product_id,m.id,SUM(v.value)/SUM(v.quantity),SUM(v.quantity),p.name, "
f"{_RECEIPT_DATE} AS day FROM {_RECEIPT_FROM} "
"WHERE m.state='done' AND src.usage='supplier' AND dst.usage='internal' AND v.quantity>0 "
"GROUP BY v.product_id,m.id,p.name,day").fetchall()}
lines = {(int(pid), int(move)): (float(amount), invalid)
for pid, move, amount, invalid in con.execute(
"SELECT a.product_id,a.move_id,COALESCE(SUM(a.additional_landed_cost/NULLIF(a.quantity,0)),0), "
"count(*) FILTER (WHERE a.quantity IS NULL OR a.quantity<=0 OR a.additional_landed_cost IS NULL) "
"FROM stock_valuation_adjustment_lines a JOIN stock_landed_cost h ON h.id=a.cost_id "
"WHERE h.state='done' AND a.product_id IS NOT NULL AND a.move_id IS NOT NULL "
"GROUP BY a.product_id,a.move_id").fetchall()}
mismatches, by_receipt = [], {}
used_total, expected_total = 0.0, 0.0
for part in usage:
pid, move, units = part["product_id"], part["move_id"], part["units"]
source = base.get((pid, move))
landed, invalid = lines.get((pid, move), (0.0, 0))
expected = landed * units
stored = saved.get(part["observation_id"])
purchase_ok = (source is not None and source[0] is not None
and abs(float(source[0]) * units - part["purchase_used"]) <= 0.005)
history_ok = stored is not None and stored == part["receipt"]
receipt_ok = (source is not None and stored is not None and source[2] == stored["receipt"]
and source[3] == stored["date"] and stored["date"] <= part["sale_date"])
ok = (history_ok and receipt_ok and purchase_ok and invalid == 0 and math.isfinite(expected)
and abs(part["landed_used"] - expected) <= 0.005)
if not ok:
mismatches.append({"sku": part["sku"], "product_id": pid, "move_id": move})
used_total += part["landed_used"]
expected_total += expected
key = pid, move
by_receipt[key] = by_receipt.get(key, 0.0) + part["units"]
overdrawn = [key for key, used in by_receipt.items() if key not in base or used > base[key][1] + 1e-8]
return [{"check": "amazon_profit: receipt COGS ties by receipt to stock_valuation_adjustment_lines",
"ok": not mismatches, "state": "verified" if not mismatches else "failed",
"a": round(used_total, 4), "b": round(expected_total, 4),
"gap": round(abs(used_total - expected_total), 6), "mismatches": mismatches},
{"check": "amazon_profit: seller SKUs share each receipt without duplicate allocation",
"ok": not overdrawn, "a": len(overdrawn), "b": 0, "gap": len(overdrawn)}]
TYPED_REFUSAL = "Not verified: typed COGS does not tie to the stored typed values."
def typed_cogs_checks(con, computed, typed, usage=None, bounds=None):
"""Typed COGS stays out of the Odoo receipt tie and ties to stored typed values times mirror units."""
typed = typed or {}
typed_skus = sorted(sku for sku, values in computed.items() if values.get("cogs_source") == COGS_TYPED)
in_tie = sorted({part.get("sku") for part in usage or ()} & set(typed_skus))
units = {item["sku"]: int(_num(item.get("units")))
for item in economics_by_sku(con, bounds["start"], bounds["end"])} if bounds else {}
share = round(sum(_num(computed[sku].get("cogs_total")) for sku in typed_skus), 2)
stored = round(sum(value * units.get(sku, 0) for sku, value in typed.items()
if (computed.get(sku) or {}).get("cogs_source") == COGS_TYPED), 2)
unstored = [sku for sku in typed_skus if sku not in typed]
tolerance = 0.005 * max(len(typed_skus), 1) + 1e-9
return [{"check": "amazon_profit: typed COGS SKUs stay out of the Odoo receipt tie",
"ok": not in_tie, "typed": typed_skus, "a": len(in_tie), "b": 0, "gap": len(in_tie)},
{"check": "amazon_profit: typed COGS share ties to stored values",
"ok": not unstored and abs(share - stored) <= tolerance, "typed": typed_skus,
"a": share, "b": stored, "gap": round(abs(share - stored), 4), "unstored": unstored}]
def stored_tables(st):
"""`{table key: (field keys, rows)}` of the three stored Command Center tables."""
import core.user_tables as ut
out = {}
for key in (TABLE_KEY, ASIN_KEY, PARENT_KEY):
table = ut.get_one(key, st) or {}
out[key] = ([f["key"] for f in table.get("fields") or []], list((table.get("rows") or {}).values()))
return out
def stored_grain_checks(st, con):
"""The stored ASIN and parent tables tie per row to the mirror over the stored period, and every
stored blank says why. Reads only."""
tables = stored_tables(st)
bounds = read_settings(st).get("lastPeriod") or {}
if not bounds.get("start") or not bounds.get("end"):
return [{"check": "amazon_profit: the stored tables name their period", "ok": False, "grain": None,
"reason": "Not verified: the stored Command Center has no period."}, blank_check(tables)]
sku_cells = {str(r.get("sku")).strip(): r for r in tables[TABLE_KEY][1] if str(r.get("sku") or "").strip()}
by_grain = {grain: {str(r.get(grain)).strip(): r for r in tables[key][1] if str(r.get(grain) or "").strip()}
for grain, key in (("asin", ASIN_KEY), ("parent_asin", PARENT_KEY))}
return grain_ties(con, bounds["start"], bounds["end"], sku_cells, by_grain) + [blank_check(tables)]
def validate(pre=None, month=None, con=None, *, cost_usage=None, cost_history=None, typed_rows=None,
typed_values=None, typed_period=None, st=None, today=None):
"""Tie net proceeds components to independent Amazon sources within 2% over settled days.
Receipt COGS ties to Odoo; typed COGS (`typed_rows`) is listed apart and ties to its stored values.
With `st`, stored grain figures, search signals and listing dates tie to their sources."""
month = month or last_complete_month()
bounds = period_bounds(month)
start, end = bounds["start"], bounds["end"]
source = "given cursor"
if con is None:
con, source = validation_cursor()
checks = []
try:
_require(con, ECONOMICS, _ECONOMICS_REQUIRED)
currency = economics_currency(con)
span = settlement_coverage(con, start, end, currency)
settled = settlements_total(con, start, end, currency)
except SourceUnavailable as exc:
label = dt.date.fromisoformat(start).strftime("%B %Y")
raise SourceUnavailable(f"Not verified: settlement components in {label}: {exc}") from None
lo, hi = span["start"] or start, span["end"] or end
proceeds = net_proceeds_total(con, lo, hi)
aligned = settlement_components(con, lo, hi, currency)
if span["start"] is None:
aligned = dict(aligned, total=None, rows=0,
reason=f"Not verified: settlement components in {bounds['label']}: no settled days.")
target = aligned["total"]
ok, gap = _tie(proceeds, target) if target is not None else (False, math.inf)
covered = {k: span[k] for k in ("start", "end", "label", "partial", "clear", "sentence")}
checks.append({"check": f"amazon_profit {month}: net proceeds tie to settlements within 2% over "
f"{span['label'] or 'no settled days'} ({gap:.2%} apart, {source})",
"ok": ok and not aligned["reason"], "a": round(proceeds, 2),
"b": None if target is None else round(target, 2), "reason": aligned["reason"],
"components": aligned["parts"], "excluded": aligned.get("excluded", {}),
"failedComponents": aligned.get("failedComponents", []),
"gap": round(gap, 4), "covered": covered, "tolerance": SETTLEMENT_TOLERANCE})
dropped = net_proceeds_total(con, lo, hi, drop_fees=True)
control_ok, control_gap = _tie(dropped, target) if target is not None else (False, math.inf)
checks.append({"check": f"amazon_profit {month}: NC net proceeds without fees miss settlements "
f"({control_gap:.2%} apart)",
"ok": not control_ok, "a": round(dropped, 2),
"b": None if target is None else round(target, 2),
"gap": round(control_gap, 4)})
checks.append({"check": f"amazon_profit {month}: every {settled['currency'] or 'settlement'} "
"settlement row carries a readable date",
"ok": settled["undated"] == 0, "a": settled["undated"], "b": 0,
"gap": settled["undated"]})
econ = economics_by_sku(con, start, end)
ours = sum(compute_row(item)["net_proceeds"] for item in econ)
reported = reported_net_proceeds(con, start, end)
checks.append({"check": f"amazon_profit {month}: per SKU net proceeds sum to Amazon's reported total",
"ok": reported is not None and abs(ours - reported) <= 0.01,
"a": round(ours, 2), "b": None if reported is None else round(reported, 2),
"gap": None if reported is None else round(abs(ours - reported), 4)})
distinct = con.execute(f"SELECT count(DISTINCT msku) FROM {ECONOMICS} "
"WHERE NULLIF(trim(msku), '') IS NOT NULL").fetchone()[0]
checks.append({"check": f"amazon_profit: every SKU is present, no cap ({len(econ)} SKUs)",
"ok": len(econ) == distinct, "a": len(econ), "b": distinct,
"gap": abs(len(econ) - distinct)})
checks.extend(landed_cost_checks(con, cost_usage, cost_history))
if typed_rows is not None:
checks.extend(typed_cogs_checks(con, typed_rows, typed_values, cost_usage, typed_period or bounds))
if st is not None:
checks.extend(stored_grain_checks(st, con))
checks.extend(listing_date_checks(st, con))
checks.extend(asin_signal_checks(st, con, today=today))
return checks
# ---------------------------------------------------------------- semantic validator
def _v_amazon_sku_net_profit(date_from, date_to, team_id=None, **_kwargs):
"""Per SKU net proceeds (the profit's revenue side) against Amazon's reported total."""
con, _source = validation_cursor()
start, end = str(date_from)[:10], str(date_to)[:10]
ours = sum(compute_row(item)["net_proceeds"] for item in economics_by_sku(con, start, end))
reported = reported_net_proceeds(con, start, end)
if reported is None:
raise SourceUnavailable("amzn_economics lacks net_proceeds_total_amount.")
return round(ours, 2), round(reported, 2)
PROFIT_VALIDATORS = {"amazon_sku_net_profit": _v_amazon_sku_net_profit}
def register_semantic_validators():
"""Register this module's validator until harness/semantic.py imports it directly."""
from harness import semantic
semantic._VALIDATORS.update(PROFIT_VALIDATORS)
return tuple(PROFIT_VALIDATORS)
try:
register_semantic_validators()
except Exception: # noqa: BLE001
pass
"""Command Center: one row per Amazon SKU with Odoo cost, profit, ads, stock, returns, health and score.
Net proceeds = net sales minus fees minus sponsored ads, from the `amzn_economics` mirror.
COGS follows incoming receipt valuations and posted landed costs through FIFO.
The module owns its table (`ut_amazon_sku_profit`). It is not a connector table, so people
edit the owned columns (Odoo product, units per SKU, owner, status, due date). A refresh
rewrites only the computed columns, keyed by SKU, so owned cells belong to the SKU and
survive every period switch.
`ut_amazon_asin_profit` and `ut_amazon_parent_profit` roll the SKU rows up on every refresh.
Projects retain their own verified period totals and Impact in the tenant projects store.
`validate()` ties each net proceeds component to independent Amazon amounts within 2% over settled
days, with order charges on purchase dates and monthly storage in its service month.
A refresh runs it for the last complete month and writes no figures when it
fails. A passing refresh stores the verified range and the day the rest is due to settle.
Owned items (R10, R31): a refresh opens one item on a Losing or Thin SKU with no status, for the
default owner, due in 14 days. It reopens a failed fix 30 days after the status was set. The
`status_set` column records the status and the date a refresh first saw it, because the
activity log is delivered async and pruned, so a refresh cannot read a status date from it.
"""
from __future__ import annotations
import atexit
import datetime as dt
import hashlib
import json
import math
import re
import shutil
import tempfile
import threading
from decimal import Decimal, ROUND_HALF_UP
from pathlib import Path
from modules import amazon_sku_metrics as M
from modules import amazon_part_dates as _part_dates
import core.user_tables as _registry
from core import module_tables
from core import amazon_rivals as _rivals
TABLE_KEY = "ut_amazon_sku_profit"
TABLE_LABEL = "Project management"
PROJECTS_KEY = "ut_amazon_projects"
PROJECTS_LABEL = "Projects"
ASIN_KEY = "ut_amazon_asin_profit"
ASIN_LABEL = "Child ASIN"
PARENT_KEY = "ut_amazon_parent_profit"
PROFIT_MAX_FIELDS = 100 # A module table declares its field cap; a connector table stays exempt.
PARENT_LABEL = "Parent ASIN"
# The four tables are the module's, not a person's: they take no MAX_TABLES slot (D-3284 spirit).
_registry.register_module_tables(TABLE_KEY, PROJECTS_KEY, ASIN_KEY, PARENT_KEY)
# S82-T70 (R33): refresh adds the SKU, ASIN and Parent ASIN rows; a person edits them but never adds
# one. Never `register_locked_records`: that turns `records_mutable` False and the edit doors then
# refuse Owner, Status, Due date and Odoo product. Projects stays open.
NO_ADD_KEYS = (TABLE_KEY, ASIN_KEY, PARENT_KEY)
NO_ADD_LINE = "Rows here come from Amazon. Edit a row instead."
module_tables.register_no_add_records(*NO_ADD_KEYS, line=NO_ADD_LINE)
SETTINGS_KEY = "amazon_profit_settings"
COST_HISTORY_KEY = "amazon_profit_cost_history"
VIA = "amazon_profit"
WRITER = "automation"
TENANT_ZERO = "royal-imports"
PERIOD_DEFAULT = "90d"
PERIOD_DAYS = {"30d": 30, "90d": 90}
THIN_DEFAULT = 0.10
STATUSES = ("Open", "Repriced", "Ads cut", "Delisted", "Keep", "COGS wrong")
#: A fix that reopens when the SKU still loses money or runs thin 30 days on. Keep never reopens.
FIX_STATUSES = ("Repriced", "Ads cut", "Delisted", "COGS wrong")
BANDS = ("Losing", "Thin")
NEEDS_REVIEW, LINKED, WRONG_MATCH = "Needs review", "Linked", "Wrong match"
#: The old names stay for callers. A decision word is a Link status option.
CONFIRMED, REJECTED = LINKED, WRONG_MATCH
MATCH_STATES = (NEEDS_REVIEW, LINKED, WRONG_MATCH)
#: Option words a stored table may still hold. `ensure_table` rewrites them, cells included.
LEGACY_MATCH_STATES = {"Confirmed": LINKED, "Rejected": WRONG_MATCH}
LINK_PURPOSE = "the Amazon SKU to its Odoo product for cost and profit"
DUE_DAYS = 14
REOPEN_DAYS = 30
TEMPLATE_KEY = "amazon_sku_profit"
ASIN_TEMPLATE_KEY = "amazon_asin_profit"
PARENT_TEMPLATE_KEY = "amazon_parent_profit"
PROJECT_STATUSES = ("Open", "Done")
PROJECT_HEALTH = ("On track", "Due this week", "Overdue", "Done")
DUE_SOON_DAYS = 7
PICKER_MONTHS = 12
SETTLEMENT_TOLERANCE = 0.02
#: Amazon releases funds about 7 days after delivery, so a month settles about two weeks after it ends.
SETTLEMENT_DEFERRAL_DAYS = 14
ECONOMICS = "amzn_economics"
SETTLEMENTS = "amzn_settlements"
LISTINGS = "amzn_listings"
LOAN_REPAYMENT = "Automated Loan Repayment"
WITHHELD_TAX = "ItemWithheldTax"
_ECONOMICS_REQUIRED = ("msku", "start_date", "sales_net_product_sales_amount",
"sales_net_units_sold", "fees_total_amount")
_SETTLEMENT_REQUIRED = ("amount", "posted_date", "transaction_type", "amount_type")