dcdpr / jp / 26606456188

66%

push

github

feat(cli, conversation): Add `--grep` and `--from`/`--until` to `c use` (#679)

`jp c use` now accepts `--grep <pattern>` and `--from`/`--until` to
narrow the picker's candidate set before the interactive prompt opens.
When the combined filter leaves exactly one conversation, it is
activated directly without prompting.

```
jp c use --grep "deployment"
jp c use --grep "rollout" --from 2w
jp c use ?p --from 3w --until 1w
```

`--grep` uses smart-case matching: all-lowercase patterns are
case-insensitive; patterns with at least one uppercase character are
case-sensitive. Searched surfaces are title, user messages, assistant
messages, reasoning, and structured output. Tool call bodies are
intentionally excluded to keep the match semantically focused on chat
content.

`--from` / `--until` compose with picker target keywords (`?`, `?p`,
`?s`, `?a`) rather than conflicting with them. This required making
`CreationRange` generic over a const `EXCLUSIVE` bool: the existing
commands (`c rm`, `c archive`) keep `CreationRange<true>` which
conflicts with the positional ID arg; `c use` uses
`CreationRange<false>` which allows both to coexist.

The shared search primitives — `ConcreteScope`, `event_scope`,
`event_lines`, `title_for`, `contains_substr`, and the new `filter_ids`
— have been extracted into `crate::shared::search` so neither `c grep`
nor `c use` has to depend on the other to get them.

The interactive picker now shows a relative timestamp column (e.g. "3
days ago") next to each conversation ID, with titles aligned across all
rows using a shared column width. The same formatting applies to the
archived picker.

---------

Signed-off-by: Jean Mertz <git@jeanmertz.com>

206 of 306 new or added lines in 5 files covered. (67.32%)

31750 of 48080 relevant lines covered (66.04%)

263.76 hits per line