Repository navigation
TABLECONTENT/TABLEGEOMETRY census allowlist is probably closable with the #80 closure measurements #82
Description
The census allowlist entry for TABLECONTENT/TABLEGEOMETRY on sample_AC1032.dwg (5 declared vs 2 walked, carried since the #75 ratchet) was justified as "not decidable from this corpus". PR #80's three class-table-free measurements make the reasoning stronger: the object stream tiles exactly (4 unclaimed prologue bytes, zero between records, zero trailing) and every hard reference resolves, so there is nowhere for three more records to sit. Apply the same owner-census argument used for DICTIONARYVAR/CELLSTYLEMAP (find the dictionary/owner that would have to hold the extra instances) and either close the allowlist entry as over-declaration or document the owner that proves otherwise.
Activity
DrunkOnJava commentedon Oct 26, 2026
On the public corpus #225 pins, TABLECONTENT and TABLEGEOMETRY are short on every ACadSharp save of the drawing from R2004 on, not only R2018: declared 9 / 8 / 7 / 6 / 5 on R2004 / R2007 / R2010 / R2013 / R2018, walked 2 on each, with the walk closed on every one (Measure run 40297255178, "Walk checks"). The declared count falls by one per release while the drawing is the same, which points at a stale registration count rather than missing records. Both, and the other shortfalls in #230, are pinned per file in
tools/corpus/baseline.tsv.