Fixes
-
Prefix matching in
_get_csv— LinkedIn appends user IDs to content CSV filenames (e.g.,Shares_26645013.csv). The parser layer now falls back to^{key}_\\d+$regex matching when exact lookup fails. Fixes 7 content parsers: posts, comments, reactions, reposts, votes, saved items, and member follows. ~950 records recovered. -
LAN ads dual-key — supports both
lan_ads_engagement(current LinkedIn export) andlinkedin_audience_network_ad_engagement(older format). 155 ad records recovered. -
Known-empty warning suppression — 4 new LinkedIn CSV types that are always empty no longer trigger WARNING noise.
Tests
- 7 new tests for prefix matching and known-empty key suppression
- 436 tests passing (ruff, pyright clean)