Blame for backend/templating.py

"""
Per-recipient message rendering: {Field} placeholders, {Field|fallback}, list fields as bullets
or an inline "A, B and C", and {if Field has "x"}...{else}...{/if} conditional blocks.

One engine serves both the real send and the Campaigns page's preview, so what is previewed is
what is sent. Row values are HTML-escaped when rendering HTML (a value containing "<" must not
become markup in someone's inbox); the subject is rendered as plain text and is not escaped.

A template that cannot be rendered safely (unclosed {if}, stray {else}, a condition that does
not parse) is reported by validate_template so a send is refused up front instead of going out
half-rendered to every recipient.
"""

from __future__ import annotations

import html as html_lib
import re
from dataclasses import dataclass, field

LIST_SPLIT_MIXED = re.compile(r"[;,|\n]")
_BLOCK_TOKEN = re.compile(r"\{(?:if\s+([^{}]*)|(else)|(/if))\}")
_PLACEHOLDER = re.compile(r"\{([A-Za-z0-9_][A-Za-z0-9_ ]*?)(?:\|([^{}]*))?\}")
_PARAGRAPH_TOKEN = re.compile(r"<p(?:\s[^>]*)?>\s*(\{(?:if\s[^{}]*|else|/if)\})\s*</p>", re.IGNORECASE)
_PARAGRAPH_LIST = re.compile(r"<p(?:\s[^>]*)?>\s*\{([A-Za-z0-9_][A-Za-z0-9_ ]*?)(?:\|(bullets|inline))?\}\s*</p>", re.IGNORECASE)
_CONDITION = re.compile(
    r"""^(?P<field>[A-Za-z0-9_][A-Za-z0-9_ ]*?)\s+
        (?P<op>is\ not\ empty|is\ empty|is\ true|is\ false|does\ not\ contain|is\ not|is|has\ not|lacks|has|contains)
        (?:\s+["“”'](?P<value>.*?)["“”'])?\s*$""",
    re.VERBOSE | re.IGNORECASE,
)
_NEEDS_VALUE = {"is", "is not", "has", "has not", "lacks", "contains", "does not contain"}

_UL_STYLE = "margin:0 0 14px;padding-left:22px;"
_LI_STYLE = "margin:0 0 4px;"


class TemplateError(ValueError):
    pass


@dataclass
class _Cond:
    field: str
    op: str
    value: str


@dataclass
class _If:
    cond: _Cond
    then: list = field(default_factory=list)
    other: list = field(default_factory=list)


def split_items(text: str, delimiter: str | None) -> list[str]:
    """Splits a list-style cell into trimmed, non-empty items. delimiter "mixed" accepts any of
    ; , | or a newline in the same cell (people are inconsistent), anything else is literal."""
    if not text:
        return []
    if delimiter == "mixed":
        parts = LIST_SPLIT_MIXED.split(text)
    else:
        parts = text.split(delimiter or ";")
    return [p.strip() for p in parts if p.strip()]


def _parse_condition(raw: str) -> _Cond:
    m = _CONDITION.match(html_lib.unescape(raw).strip())
    if not m:
        raise TemplateError(f'Could not read the condition "{{if {raw.strip()}}}"')
    op = re.sub(r"\s+", " ", m.group("op").lower())
    value = m.group("value")
    if op in _NEEDS_VALUE and value is None:
        raise TemplateError(f'The condition "{{if {raw.strip()}}}" needs a value in quotes')
    return _Cond(m.group("field").strip(), op, (value or "").strip())


def _parse(template: str) -> list:
    """Template text -> tree of str and _If nodes. Raises TemplateError on unbalanced blocks."""
    root: list = []
    stack: list[tuple[_If, bool]] = []  # (node, currently_in_else)
    current = root
    pos = 0
    for m in _BLOCK_TOKEN.finditer(template):
        if m.start() > pos:
            current.append(template[pos:m.start()])
        pos = m.end()
        if m.group(1) is not None:
            node = _If(_parse_condition(m.group(1)))
            current.append(node)
            stack.append((node, False))
            current = node.then
        elif m.group(2):
            if not stack or stack[-1][1]:
                raise TemplateError("Found {else} without a matching {if ...}")
            node, _ = stack[-1]
            stack[-1] = (node, True)
            current = node.other
        else:
            if not stack:
                raise TemplateError("Found {/if} without a matching {if ...}")
            stack.pop()
            current = stack[-1][0].other if stack and stack[-1][1] else (stack[-1][0].then if stack else root)
    if stack:
        raise TemplateError("An {if ...} block is missing its {/if}")
    if pos < len(template):
        current.append(template[pos:])
    return root


def _column_defs(columns: list[dict] | None) -> dict[str, dict]:
    return {c["name"]: c for c in (columns or [])}


def _evaluate(cond: _Cond, row: dict, cols: dict[str, dict]) -> bool:
    col = cols.get(cond.field)
    if col is None and cols:
        return False  # a condition on a field the audience does not have never matches
    text = str(row.get(cond.field, "") or "")
    col_type = (col or {}).get("type", "text")
    needle = cond.value.strip().lower()
    op = cond.op

    if op == "is empty":
        return text.strip() == ""
    if op == "is not empty":
        return text.strip() != ""
    if op in ("is true", "is false"):
        truthy = text.strip().lower() in {"true", "yes", "y", "1"}
        return truthy if op == "is true" else not truthy

    if col_type == "tag":
        items = {i.lower() for i in split_items(text, col.get("delimiter"))}
        if op in ("has", "is"):
            return needle in items
        if op in ("lacks", "has not", "is not"):
            return needle not in items
    hay = text.strip().lower()
    if op == "is":
        return hay == needle
    if op == "is not":
        return hay != needle
    if op in ("has", "lacks", "has not"):
        # On a plain column "has" means the whole word/phrase appears, so "yes" is not
        # matched by "yesterday".
        found = re.search(r"(?<!\w)" + re.escape(needle) + r"(?!\w)", hay) is not None
        return found if op == "has" else not found
    if op == "contains":
        return needle in hay
    if op == "does not contain":
        return needle not in hay
    return False


def _inline_join(items: list[str]) -> str:
    if len(items) <= 1:
        return "".join(items)
    return ", ".join(items[:-1]) + " and " + items[-1]


def _bullets(items: list[str], escape: bool) -> str:
    esc = html_lib.escape if escape else (lambda s: s)
    lis = "".join(f'<li style="{_LI_STYLE}">{esc(i)}</li>' for i in items)
    return f'<ul style="{_UL_STYLE}">{lis}</ul>'


def _render_text(text: str, row: dict, cols: dict[str, dict], html: bool) -> str:
    esc = html_lib.escape if html else (lambda s: s)

    if html:
        # A list field alone in its paragraph becomes a real bullet list (replacing the paragraph).
        def paragraph_list(m: re.Match) -> str:
            name, mode = m.group(1).strip(), (m.group(2) or "").lower()
            col = cols.get(name)
            if not col or col.get("type") != "tag" or mode == "inline":
                return m.group(0)
            items = split_items(str(row.get(name, "") or ""), col.get("delimiter"))
            return _bullets(items, True) if items else ""

        text = _PARAGRAPH_LIST.sub(paragraph_list, text)

    def placeholder(m: re.Match) -> str:
        name, option = m.group(1).strip(), m.group(2)
        col = cols.get(name)
        raw = str(row.get(name, "") or "")
        if col and col.get("type") == "tag":
            items = split_items(raw, col.get("delimiter"))
            if items:
                if option and option.lower() == "bullets" and html:
                    return _bullets(items, True)
                return esc(_inline_join(items))
        elif raw.strip():
            return esc(raw)
        if option and option.lower() not in ("bullets", "inline"):
            return option  # fallback text written by the author
        return ""

    return _PLACEHOLDER.sub(placeholder, text)


def _render(nodes: list, row: dict, cols: dict[str, dict], html: bool) -> str:
    out = []
    for node in nodes:
        if isinstance(node, str):
            out.append(_render_text(node, row, cols, html))
        else:
            branch = node.then if _evaluate(node.cond, row, cols) else node.other
            out.append(_render(branch, row, cols, html))
    return "".join(out)


def _prepare(template: str) -> str:
    # {if}/{else}/{/if} typed on their own line become their own <p>; drop that wrapper so the
    # block leaves no empty paragraphs behind.
    return _PARAGRAPH_TOKEN.sub(r"\1", template)


def validate_template(template: str, columns: list[dict] | None = None) -> list[str]:
    """Problems that would make the template render wrongly. Empty list means safe to send."""
    problems: list[str] = []
    try:
        tree = _parse(_prepare(template))
    except TemplateError as exc:
        return [str(exc)]
    names = {c["name"] for c in (columns or [])}
    if names:
        def walk(nodes: list) -> None:
            for n in nodes:
                if isinstance(n, _If):
                    if n.cond.field not in names:
                        problems.append(f'A condition uses the field "{n.cond.field}", which this audience does not have')
                    walk(n.then)
                    walk(n.other)
        walk(tree)
    return list(dict.fromkeys(problems))


def render(template: str, row: dict, columns: list[dict] | None = None, html: bool = True) -> str:
    """Renders one recipient's copy. Raises TemplateError if the template is malformed."""
    tree = _parse(_prepare(template) if html else template)
    return _render(tree, row, _column_defs(columns), html)