feat: allow for format 115 #4

Activity

1. korenmiklos

   added a commit that references this issue on Jul 7, 2026

   Add read support for legacy formats 113-115

   ...

   a9f46d4
2. korenmiklos commented on Jul 7, 2026

   

   korenmiklos

   on Jul 7, 2026 – with Claude

   MemberAuthor

   More actions

   Read support implemented on branch claude/security-bugs-performance-review-qdmmco in commit a9f46d4 — and since formats 113–115 share one layout, this covers 113 (Stata 8/9) and 114 (Stata 10/11) as well as 115 (Stata 12).

   The reader detects the format from the first byte (< → XML-tagged 117+ parser, 113/114/115 → new legacy parser for the flat binary layout: fixed header, descriptors, expansion fields). Legacy one-byte type codes are translated to the modern ones at parse time, so everything downstream — missing-value handling, %td date conversion, Latin-1→UTF-8 transcoding, value labels, projection/filter pushdown, parallel scans — works identically on legacy files. Value-label tables (same layout as 117 minus the tags) parse through a now-shared code path.

   Fixtures and tests: format_114.dta and value_labels_114.dta written by pandas; format_115.dta is a byte-identical 114-layout file with the format byte set to 115 (pandas can't write 115) carrying Latin-1 content (café, naïve) to exercise transcoding. Suite is green at 625 assertions.

   Scope note: this is read-only, as requested — COPY ... (FORMAT dta) still writes format 119. If write support for older formats is wanted, that's the same work item as the 117-write half of #3 (version-parameterized writer metadata). Also not covered: formats older than 113 (pre-Stata 8), which use different descriptor/value-label layouts.

   ---

   Generated by Claude Code