wikivoyage: drop lone file-link option lines (v2#116), rebuild four table rows, relabel (v2#110) (#133)

- wikivoyage: drop lone file-link option lines (v2#116), rebuild four table rows, relabel (v2#110) (d65ffc239db02e33f7034be31e1496dd7fd242f8)

artifacts/source_findings.md

@@ -2877,3 +2877,30 @@ characters, +214 tokens, labels unchanged. In wikivoyage and wikibooks the new r
 
 **Blockers** — none to merge. Until the wikivoyage and wikibooks releases are cleaned with the same
 rule (their own PRs, next), their contract tests fail on `drop_residue(text) == text`.
+
+## wikivoyage — lone option lines removed, four table rows rebuilt (update to the sections above)
+
+**Investigated** — 2026-10-18 (slayerlabs/polish-dynaword-v2#116: the rule of the wikinews section
+above, applied to this release).
+
+**Verdict** — `kept`, text cleaned again and relabelled: 13,640 documents, 44,720,085 characters,
+16,959,169 tokens (was 44,719,971 / 16,959,180). Labels per record unchanged in count:
+`CC-BY-SA-3.0` 11,220 documents (6,163,588 tokens, 36.3%), `CC-BY-SA-3.0 AND CC-BY-SA-4.0` 1,672
+(9,834,703, 58.0%), `CC-BY-SA-4.0` 748 (960,878, 5.7%).
+
+**Effect** — 18 rows lose 24 lines (`mały` 23, `right` 1; in 6 rows the first line), 124 characters
+and 72 tokens. One of them (`wikivoyage_5942`) had no other residue, so the cleanup now changed
+13,172 rows. No page fell under 200 characters. The rerun started from the data before Hub PR 125
+(commit `ae3214b`) and ran `retext`, then `wiki_history.py` `scan`, `fetch` and `compare` on the same
+stub dump (sha1 `6ca8ea9e…`), `relabel` and `datasheet`. The labels file differs in three rows (`wikivoyage_2833`, `5396`, `15013`), each counting one added word
+fewer; all three stay `CC-BY-SA-3.0 AND CC-BY-SA-4.0`.
+
+**Table rows rebuilt** — the rebuild also carries the table-start fix of the wikibooks section above
+to the four rows released with a lone `{` (`wikivoyage_2055`, `2564`, `9129`, `10490`): each `{`
+line is again the table start `{| class="wikitable"` (in `10490` `prettytable`, in `2564` with
+`mw-collapsible mw-collapsed`), adding 238 characters and 61 tokens. With the previous rule the
+chain gives main except these four rows, which the main code builds this way too.
+`build.json` gains an empty `dropped_duplicate` list, which `retext` has written since Hub PR 130.
+
+**Blockers** — none to merge. Until the wikibooks release is cleaned with the same rule (its own PR,
+next), its contract test fails on `drop_residue(text) == text`.

data/wikivoyage/wikivoyage.build.json

@@ -63,15 +63,16 @@
   },
   "residue_cleanup": {
     "after": {
-      "chars": 44719971,
+      "chars": 44720085,
       "docs": 13640,
-      "tokens": 16959180
+      "tokens": 16959169
     },
     "before": {
       "chars": 45176480,
       "docs": 13643,
       "tokens": 17125240
     },
+    "dropped_duplicate": [],
     "dropped_short": [
       {
         "chars": 188,
@@ -90,7 +91,7 @@
       }
     ],
     "issue": "slayerlabs/polish-dynaword-v2#116",
-    "rows_changed": 13171,
+    "rows_changed": 13172,
     "step": "fetch_wikimedia.drop_residue"
   },
   "rightsinfo": {

data/wikivoyage/wikivoyage.license-evidence.json

@@ -40,9 +40,9 @@
       "tokens": 0
     },
     "tokens": {
-      "CC-BY-SA-3.0": 6163617,
-      "CC-BY-SA-3.0 AND CC-BY-SA-4.0": 9834655,
-      "CC-BY-SA-4.0": 960908
+      "CC-BY-SA-3.0": 6163588,
+      "CC-BY-SA-3.0 AND CC-BY-SA-4.0": 9834703,
+      "CC-BY-SA-4.0": 960878
     }
   },
   "review_scope": "licence of contributors' text on pl.wikivoyage.org and its version per contribution; not independent legal advice",

data/wikivoyage/wikivoyage.md

@@ -17,9 +17,9 @@ Each row is labelled by the contributions its text holds:
 
 | `license` | documents | tokens | share of tokens |
 |---|---:|---:|---:|
-| `CC-BY-SA-3.0` | 11,220 | 6,163,617 | 36.3% |
-| `CC-BY-SA-3.0 AND CC-BY-SA-4.0` | 1,672 | 9,834,655 | 58.0% |
-| `CC-BY-SA-4.0` | 748 | 960,908 | 5.7% |
+| `CC-BY-SA-3.0` | 11,220 | 6,163,588 | 36.3% |
+| `CC-BY-SA-3.0 AND CC-BY-SA-4.0` | 1,672 | 9,834,703 | 58.0% |
+| `CC-BY-SA-4.0` | 748 | 960,878 | 5.7% |
 
 A row labelled `CC-BY-SA-3.0 AND CC-BY-SA-4.0` holds text under each licence. It is not split by contribution;
 its page history tells the two apart. The basis of each label is the sidecar's `license_basis`:
@@ -78,8 +78,8 @@ Articles (namespace 0) only. Redirects, empty pages and pages shorter than 200 c
 Wikitext is converted with mwparserfromhell `strip_code(normalize=True, collapse=True)`, as in v0.2.5,
 then `drop_residue` removes what that leaves of category and file links: `Kategoria:...` lines, bare file
 titles and file-link options (`thumb|250px|...`); captions stay ([v2#116](https://github.com/slayerlabs/polish-dynaword-v2/issues/116)).
-It was applied to the committed text (`fetch_wikimedia.py retext`): 13,171 rows changed,
-456,509 characters and 166,060 tokens removed, and 3 pages that fell under 200
+It was applied to the committed text (`fetch_wikimedia.py retext`): 13,172 rows changed,
+456,395 characters and 166,071 tokens removed, and 3 pages that fell under 200
 characters were dropped (ids and titles in `wikivoyage.build.json`). Templates and comments are dropped; `<ref>`
 bodies stay in the text (the wikipedia fetcher's `strip_article` removes them). Of two pages with the same text
 (sha1 of the stripped text), the lower page id is kept.
@@ -89,7 +89,7 @@ bodies stay in the text (the wikipedia fetcher's `strip_article` removes them).
 ## Statistics
 | documents | characters | tokens (tiktoken proxy) |
 |---:|---:|---:|
-| 13,640 | 44,719,971 | 16,959,180 |
+| 13,640 | 44,720,085 | 16,959,169 |
 
 v0.2.5 had 13,645 documents, 45,184,300 characters and 17,128,200 tokens; the difference is the 2 pages withheld by the rights review and the residue cleanup (see Provenance).
 
@@ -118,8 +118,8 @@ Llama-3 count is computed at release.
   `CC-BY-SA-3.0 AND CC-BY-SA-4.0`, with no threshold
 - text moved or merged from an older page into a page created after the cutoff (labelled `CC-BY-SA-4.0` alone)
 - the time of day of the switch (see Licensing)
-- what `drop_residue` leaves: file-link options glued to the word before them (`Ladadikathumb|...`), a line
-  holding only `mały` (an option, but also a Polish word), category links inside a line of prose, and `<ref>` bodies
+- what `drop_residue` leaves: file-link options glued to the word before them (`Ladadikathumb|...`), category
+  links inside a line of prose, and `<ref>` bodies
 - text imported from elsewhere under another licence or with extra attribution requirements
 - text first contributed under GFDL before the 2009 move to CC BY-SA, and GFDL dual
   licensing; only the CC BY-SA grant is relied on

data/wikivoyage/wikivoyage.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1436ad45f3d86163537cfa8f2c1f642554440953935a9bc812bca9d7bdc85da7
-size 19846346
+oid sha256:fdb4233d73af96b5370e46c8c56102081b38c388be03c58858baa149f8c3aeb7
+size 19846184

data/wikivoyage/wikivoyage.stats.json

@@ -5,8 +5,8 @@
   "drop_lang": 9,
   "drop_dup": 0,
   "drop_ocr": 0,
-  "chars": 44719971,
-  "tokens": 16959180,
+  "chars": 44720085,
+  "tokens": 16959169,
   "licenses": {
     "CC-BY-SA-3.0": 11220,
     "CC-BY-SA-3.0 AND CC-BY-SA-4.0": 1672,

src/fetch_wikimedia.py

@@ -849,8 +849,8 @@ NOT_CHECKED_BY_CONTRIBUTION = {
   `CC-BY-SA-3.0 AND CC-BY-SA-4.0`, with no threshold
 - text moved or merged from an older page into a page created after the cutoff (labelled `CC-BY-SA-4.0` alone)
 - the time of day of the switch (see Licensing)
-- what `drop_residue` leaves: file-link options glued to the word before them (`Ladadikathumb|...`), a line
-  holding only `mały` (an option, but also a Polish word), category links inside a line of prose, and `<ref>` bodies
+- what `drop_residue` leaves: file-link options glued to the word before them (`Ladadikathumb|...`), category
+  links inside a line of prose, and `<ref>` bodies
 - text imported from elsewhere under another licence or with extra attribution requirements
 - text first contributed under GFDL before the 2009 move to CC BY-SA, and GFDL dual
   licensing; only the CC BY-SA grant is relied on

src/test_wikivoyage_contract.py

@@ -13,15 +13,20 @@ WITHHELD = (2, 7820, 2960)
 
 def test_wikivoyage_contract(tmp_path):
     stats, ev = check_released("wikivoyage", tmp_path, dropped=WITHHELD)
-    assert (stats["kept"], stats["tokens"]) == (13640, 16959180)
-    # v2#116: drop_residue changed 13,171 rows; 3 pages fell under 200 characters
+    assert (stats["kept"], stats["tokens"]) == (13640, 16959169)
+    # v2#116: drop_residue changed 13,172 rows; 3 pages fell under 200 characters
     build = json.loads((ROOT / "data" / "wikivoyage" / "wikivoyage.build.json").read_text(encoding="utf-8"))
     clean = build["residue_cleanup"]
-    assert clean["rows_changed"] == 13171 and clean["before"] == {"docs": 13643, "chars": 45176480, "tokens": 17125240}
+    assert clean["rows_changed"] == 13172 and clean["before"] == {"docs": 13643, "chars": 45176480, "tokens": 17125240}
     assert [r["id"] for r in clean["dropped_short"]] == ["wikivoyage_11830", "wikivoyage_14345", "wikivoyage_15641"]
+    # the rows where the table-start fault fixed in Hub PR 127 had left a lone "{" of "{| class=..." are rebuilt
+    texts = dict(zip(*pq.read_table(ROOT / "data" / "wikivoyage" / "wikivoyage.parquet", columns=["id", "text"]).to_pydict().values()))
+    for i in (2055, 2564, 9129, 10490):
+        lines = texts[f"wikivoyage_{i}"].split("\n")
+        assert "{" not in lines and any(l.startswith("{| class=") for l in lines), i
     # labelled by contribution (v2#110): 3.0, 4.0 or both
     assert stats["licenses"] == {"CC-BY-SA-3.0": 11220, "CC-BY-SA-3.0 AND CC-BY-SA-4.0": 1672, "CC-BY-SA-4.0": 748}
-    assert ev["release_scope"]["tokens"] == {"CC-BY-SA-3.0": 6163617, "CC-BY-SA-3.0 AND CC-BY-SA-4.0": 9834655, "CC-BY-SA-4.0": 960908}
+    assert ev["release_scope"]["tokens"] == {"CC-BY-SA-3.0": 6163588, "CC-BY-SA-3.0 AND CC-BY-SA-4.0": 9834703, "CC-BY-SA-4.0": 960878}
     labels = build["licence_labels"]
     assert labels["cases"] == {"shipped_before_cutoff": 9039, "same_revision_text": 37, "identical_text": 1976,
                                "no_text_added": 168, "created_after_cutoff": 748, "changed_after_cutoff": 1672}