add info about full preproc

README.md

@@ -26,6 +26,20 @@ text, wikitext, direct categories, community quality status and revision links,
 and adds a `topics` list. The `topics` configuration supplies the complete subject
 definitions. The full category catalog remains in the source collection.
 
+## Files at a glance
+
+| File | Purpose | Bytes |
+| --- | --- | ---: |
+| `data/articles.jsonl.gz` | Source articles with retained subject labels. | 413,342,335 |
+| `data/topics.jsonl.gz` | The 28 subject definitions and their boundaries. | 2,671 |
+| `data/labels.json` | Article-ID to subject-slug mapping for replay and analysis. | 649,823 |
+| `data/noow.csv` | Whole-document text and ordered normalized VW features. | 1,118,455,706 |
+| `data/bow.csv` | The same documents with counted VW features. | 875,143,956 |
+
+[SHA256SUMS](SHA256SUMS) covers this repository's published data. Preparation code,
+environment locks and source/output pins are maintained once in the
+[Full collection](https://huggingface.co/datasets/TopicNet/RuWiki-Good-2026-V2/tree/main/preprocessing).
+
 ## Selection and subject labels
 
 The source includes namespace-0 articles with active recommended (Добротные),
@@ -62,7 +76,7 @@ research train/test split. The `corpus` split is a storage partition. Source tex
 has not been shortened or rewritten for subject selection. No exact duplicate source text was
 removed. This collection is not a representative sample of Russian writing.
 
-## Loading
+## Quick start: load the published files
 
 The `articles` configuration is the default. `articles` and `topics` use
 compressed JSON Lines and the `corpus` split, with no dataset script.
@@ -80,8 +94,7 @@ print(len(articles))  # 17251
 print(len(topics))    # 28
 ```
 
-From a downloaded repository, use its local directory with the same two config
-names, or load the files directly:
+From a downloaded repository, load the JSON Lines files directly:
 
 ```python
 articles = load_dataset(
@@ -121,6 +134,31 @@ bow = load_dataset(
 )
 ```
 
+CSV is the primary prepared format. The explicit `"csv"` builder above is
+required: these files are not additional YAML configurations, because combining
+JSON and CSV builders in one repository configuration did not load reliably.
+The published CSV files have been fully read with Hugging Face Datasets.
+
+To export one VW document per line without loading the CSV into memory, run this
+from the downloaded repository. Choose an output path outside the repository:
+
+```python
+import csv
+from pathlib import Path
+
+csv.field_size_limit(2**31 - 1)
+with Path("data/noow.csv").open(encoding="utf-8", newline="") as source:
+    with Path("../RuWiki-Good-2026-V2-Selected-noow.vw").open("w", encoding="utf-8", newline="") as output:
+        for row in csv.DictReader(source):
+            output.write(row["vw_text"] + "\n")
+```
+
+The field-size setting is needed only for Python's standard `csv` reader;
+it does not configure Hugging Face Datasets. Use `bow.csv` in the same example
+for counted features. CSV records contain whole documents and may include
+embedded newlines in `raw_text`, so splitting the CSV by physical lines is unsafe.
+The repository does not store a second permanent copy as VW text files.
+
 The NLP preparation section describes the normalized namespaces and text fields.
 
 The sidecar [data/labels.json](data/labels.json) is a JSON object mapping every
@@ -244,18 +282,10 @@ Counts overlap because an article may have several subjects.
 
 
 The source is the [Russian Wikipedia 20261001 dump](https://dumps.wikimedia.org/ruwiki/20261001/).
-The exact source files are listed below with their SHA-256 digests. Dump mirrors
-may later remove this dated snapshot; the filenames and digests identify the
-inputs independently of their continued availability.
-
-| File | SHA-256 |
-| --- | --- |
-| `ruwiki-20261001-page.sql.gz` | `24250cca6122fff02896a1ced0a85b7afa4d6be00fa119c169a5eeef3fbe5830` |
-| `ruwiki-20261001-page_props.sql.gz` | `fc9d5aeaf3487d932b82d18a16bc96f0e4e4702e1149cbbdc482eb509f6deb60` |
-| `ruwiki-20261001-categorylinks.sql.gz` | `1ecb32ffaff28a120b6a8f4297d4b228b65c49beb593eae022b5811ae41512e1` |
-| `ruwiki-20261001-linktarget.sql.gz` | `bc1f24b7b8eb06298cbe70fdfb010ec023586d05bd2d3403ba2370091e1edc4a` |
-| `ruwiki-20261001-pages-articles-multistream.xml.bz2` | `c22b6441176d2bc572ed6883abc6292fe168efca211f6cdd61d8772a78492a1b` |
-| `ruwiki-20261001-pages-articles-multistream-index.txt.bz2` | `17c6a4cebb90ecb2b0f881e04aa3b49356bd6ee6add802a7201b9ca1a9b093e3` |
+The [source and output pins](https://huggingface.co/datasets/TopicNet/RuWiki-Good-2026-V2/blob/main/preprocessing/wikipedia-quality-corpus-20261001.json) record the names, byte sizes and SHA-256 digests of all six
+inputs: page, page properties, category links and link targets SQL dumps, plus the
+multistream article XML and its index. Dump mirrors may later remove this dated
+snapshot; the pins identify the inputs independently of their availability.
 
 SQL provides selection, titles, page properties and category relationships. XML
 provides article content, `revision_id` and `revision_timestamp`. In the 17,413-article source collection, **11 articles**
@@ -275,18 +305,37 @@ can inspect these losses or apply a different extraction procedure. Neither
 
 The [standalone preparation code](https://huggingface.co/datasets/TopicNet/RuWiki-Good-2026-V2/tree/main/preprocessing)
 and [full instructions](https://huggingface.co/datasets/TopicNet/RuWiki-Good-2026-V2#nlp-preparation)
-are maintained in the source collection. Download the code directory and use
-`--dataset selected` to prepare this collection, or `--dataset both` to prepare
-both collections from their pinned sources. Build and verify require Python
-3.11.15; verify needs no NLP packages.
+are maintained in Full. Install that directory's locked Linux x86-64 environment
+before running the commands below from the parent of both downloaded repository
+directories. Working files go under a separate `ruwiki-rebuild` directory; choose
+fresh output and scratch paths on a disk with enough space. Corpus output and
+scratch inside a published clone are rejected. Build and
+verify require Python 3.11.15 because VW escaping depends on its Unicode database;
+verify needs no NLP packages. `requirements.txt` includes `mwparserfromhell==0.7.2`
+for corpus extraction; the existing environment locks also remain available.
 
 ```bash
-python prepare.py --action download --dataset selected --sources-dir ./sources
-python prepare.py --action build --dataset selected --sources-dir ./sources --output-dir ./output --scratch-dir ./scratch
-python prepare.py --action verify --dataset selected --sources-dir ./sources --output-dir ./output
+conda create --name ruwiki-nlp --file RuWiki-Good-2026-V2/preprocessing/conda-linux-64.lock.txt
+conda activate ruwiki-nlp
+python -m pip install -r RuWiki-Good-2026-V2/preprocessing/pip-linux-x86_64.lock.txt
+python RuWiki-Good-2026-V2/preprocessing/prepare.py --stage nlp --action download --dataset selected --sources-dir ./ruwiki-rebuild/nlp-sources
+python RuWiki-Good-2026-V2/preprocessing/prepare.py --stage nlp --action build --dataset selected --sources-dir ./ruwiki-rebuild/nlp-sources --output-dir ./ruwiki-rebuild/nlp-output --scratch-dir ./ruwiki-rebuild/nlp-scratch
+python RuWiki-Good-2026-V2/preprocessing/prepare.py --stage nlp --action verify --dataset selected --sources-dir ./ruwiki-rebuild/nlp-sources --output-dir ./ruwiki-rebuild/nlp-output
 ```
 
-The workflow produces `output/selected/noow.csv` and `bow.csv` with exactly
+`--stage nlp` is the default; commands without it continue to work. Use
+`--dataset both` to prepare both collections. Download checks pinned source
+snapshots and supplies the pinned NLTK stopwords archive. Build runs offline.
+Keep output and scratch separate from published `data/` and use fresh directories
+for a new build. Full's instructions explain reuse of local JSONL files and
+access to restricted source repositories.
+
+After the structural NLP checks, the shared entry point independently compares
+CSV and Selected label byte sizes and SHA-256 hashes with published expectations
+in the corpus contract, reporting `PUBLISHED_NLP_OK`. It does not rely on the
+new build manifest as its authority and does not replay NLP normalization.
+
+The workflow produces `./ruwiki-rebuild/nlp-output/selected/noow.csv` and `bow.csv` with exactly
 `id`, `raw_text`, `vw_text` columns for all 17,251 selected IDs, preserving article
 order. CSV `raw_text` is the title, two LF characters and the original readable
 body with its leading CR/LF characters removed. Original JSON `raw_text` remains
@@ -321,13 +370,63 @@ Load a generated CSV directly:
 
 ```python
 noow = load_dataset(
-    "csv", data_files={"corpus": "output/selected/noow.csv"}, split="corpus"
+    "csv", data_files={"corpus": "./ruwiki-rebuild/nlp-output/selected/noow.csv"}, split="corpus"
 )
 bow = load_dataset(
-    "csv", data_files={"corpus": "output/selected/bow.csv"}, split="corpus"
+    "csv", data_files={"corpus": "./ruwiki-rebuild/nlp-output/selected/bow.csv"}, split="corpus"
 )
 ```
 
+A measured NLP build of **both Full and Selected** from ready JSONL snapshots took
+42 min 5.83 s, with 10.58 GiB peak resident memory, approximately 3.74 GiB of final
+output and 0.53 GiB of scratch. These are measurements from one environment,
+not Selected-only estimates or measurements of replay from raw Wikimedia dumps.
+
+## Rebuild Selected from source data
+
+Full's [corpus replay instructions](https://huggingface.co/datasets/TopicNet/RuWiki-Good-2026-V2#rebuild-the-corpus-from-wikimedia-dumps)
+cover the six pinned 20261001 dumps, explicit download resumption with `--resume`,
+offline construction and independent published-payload verification. The shared
+[source and output contract](https://huggingface.co/datasets/TopicNet/RuWiki-Good-2026-V2/blob/main/preprocessing/wikipedia-quality-corpus-20261001.json)
+contains source dump and annotation pins. The new public path from raw dumps has
+**not yet been run end to end**; its runtime and resource requirements remain
+unmeasured.
+
+For the complete path, run from the parent of both downloaded repository
+directories, as in the NLP commands above:
+
+```bash
+python RuWiki-Good-2026-V2/preprocessing/prepare.py --stage corpus --action download --dataset both --sources-dir ./ruwiki-rebuild/dump-sources
+# Add --resume explicitly when continuing an interrupted download.
+python RuWiki-Good-2026-V2/preprocessing/prepare.py --stage corpus --action build --dataset both --sources-dir ./ruwiki-rebuild/dump-sources --annotations-dir RuWiki-Good-2026-V2-Selected/data --output-dir ./ruwiki-rebuild/corpus-output --scratch-dir ./ruwiki-rebuild/corpus-scratch
+python RuWiki-Good-2026-V2/preprocessing/prepare.py --stage corpus --action verify --dataset both --sources-dir ./ruwiki-rebuild/dump-sources --output-dir ./ruwiki-rebuild/corpus-output
+```
+
+This writes `./ruwiki-rebuild/corpus-output/full/articles.jsonl.gz` and `categories.jsonl.gz`,
+plus `./ruwiki-rebuild/corpus-output/selected/articles.jsonl.gz` and `topics.jsonl.gz`. The
+annotation directory must already contain this collection's `labels.json` and
+`topics.jsonl.gz`. Reconstruction retains the published mapping and does not
+rerun AI annotation or establish new evidence for label correctness.
+
+If a previously accepted Full corpus is already available, build only Selected
+without rereading the dumps:
+
+```bash
+python RuWiki-Good-2026-V2/preprocessing/prepare.py --stage corpus --action build --dataset selected --sources-dir RuWiki-Good-2026-V2/data --full-dir RuWiki-Good-2026-V2/data --annotations-dir RuWiki-Good-2026-V2-Selected/data --output-dir ./ruwiki-rebuild/selected-output --scratch-dir ./ruwiki-rebuild/selected-scratch
+python RuWiki-Good-2026-V2/preprocessing/prepare.py --stage corpus --action verify --dataset selected --sources-dir RuWiki-Good-2026-V2/data --output-dir ./ruwiki-rebuild/selected-output
+```
+
+This Selected-only workflow has been verified byte for byte against the published
+archive of all 17,251 articles, the labels file and the topic definitions.
+
+Use fresh output and scratch directories. Corpus verification checks exact
+compressed output bytes against the separately bundled published payload pins;
+it does not treat a newly generated manifest as independent evidence. To prepare
+NLP CSVs from `./ruwiki-rebuild/corpus-output`, run the earlier NLP commands with
+`--sources-dir ./ruwiki-rebuild/corpus-output` and fresh `--output-dir ./ruwiki-rebuild/replay-nlp-output` and
+`--scratch-dir ./ruwiki-rebuild/replay-nlp-scratch`. Run NLP download first to supply NLTK
+stopwords, then offline build and verify.
+
 Preparation code is MIT licensed in the source collection. Article text and its
 normalized derivatives retain the CC BY-SA 4.0 terms and attribution obligations
 below.
@@ -356,15 +455,17 @@ offensive material. The partial subject labels described here do not establish a
 annotation. Further downstream preprocessing, deduplication policy and research
 splits must be specified separately.
 
-## Published file identity
+## File verification
 
-| File | Bytes | SHA-256 |
-| --- | ---: | --- |
-| `data/articles.jsonl.gz` | 413,342,335 | `9eb7efffe99fc7c9bc3f4b3eb9f0378d91f387081f97cf9abea9c1686797c1ed` |
-| `data/topics.jsonl.gz` | 2,671 | `c8cf602a924f18255e155cdcaa9357907461851e849113c72051c2b89484f22c` |
-| `data/noow.csv` | 1,118,455,706 | `2c71aa6b5a327ebe924f5f0160eaf6f923a86da1a3a7fe8c68aa67f8858f8b6b` |
-| `data/bow.csv` | 875,143,956 | `bff3a82c29ab3b866fe86135ff99b3897deb98dc4e5ed6db4ea7b24bd4011bae` |
-| `data/labels.json` | 649,823 | `e7ceacd69d76601df24bb42e6d0177f46142d0b8ac1bf6e8c22ad4d5a0b0ca70` |
+Run from this downloaded repository's root:
+
+```bash
+sha256sum -c SHA256SUMS
+```
 
-The source article payload SHA-256 is
-`c51f9b6bee4c85f25ffb861060f52490a9276bae289d2aa2db7e0d10e2dda950`.
+[SHA256SUMS](SHA256SUMS) covers all five data files, excluding this README and
+Git/service metadata. Full has its own manifest for its data and shared
+preparation files. The shared corpus contract independently pins expected
+published reconstruction payloads. A successful local checksum check establishes
+agreement with the supplied manifest; use a trusted repository revision to
+establish provenance.

SHA256SUMS

@@ -0,0 +1,5 @@
+9eb7efffe99fc7c9bc3f4b3eb9f0378d91f387081f97cf9abea9c1686797c1ed  data/articles.jsonl.gz
+bff3a82c29ab3b866fe86135ff99b3897deb98dc4e5ed6db4ea7b24bd4011bae  data/bow.csv
+e7ceacd69d76601df24bb42e6d0177f46142d0b8ac1bf6e8c22ad4d5a0b0ca70  data/labels.json
+2c71aa6b5a327ebe924f5f0160eaf6f923a86da1a3a7fe8c68aa67f8858f8b6b  data/noow.csv
+c8cf602a924f18255e155cdcaa9357907461851e849113c72051c2b89484f22c  data/topics.jsonl.gz