Goal: raise information density of v2ex topic <id> --with-replies text output. Real-world example: topic 1213663 (公益站 ID 投递) has ~340 replies, of which the vast majority are 'ID: NNNN' promotional posts or 'mark/+1' noise — the actual discussion drowns.
Design:
- exact-duplicate group → one row + (× N similar)
- ID-only replies (regex ^\sID[:：]?\s\d+) → trailing single-line bucket
- short / stop-listed noise (mark, +1, 顶, 感谢, 👍, ...) → trailing bucket
- --json output stays raw (agents re-derive)
- --no-collapse opts back to current behavior
Goal: raise information density of
v2ex topic <id> --with-repliestext output. Real-world example: topic 1213663 (公益站 ID 投递) has ~340 replies, of which the vast majority are 'ID: NNNN' promotional posts or 'mark/+1' noise — the actual discussion drowns.Design: