Module search 

Recursive search for datasets below a directory.

The home screen’s filter is a fuzzy match over rows that are already listed. This module is what puts more rows in front of it: one bounded walk of the working directory, run once in the background, whose result is then filtered in memory like everything else. Nothing here is repeated per keystroke — a walk per character is how a file finder becomes slow on exactly the trees where it matters.

Every limit exists because some real directory violates it. See [Limits].

The walk keeps every data file it finds, and the filter is scored against that index off the UI thread (score). The cap on what is listed counts matches, not files: a file that matches is never lost behind thousands that do not.

Structs

Matches

What the filter matched among the files a walk kept.

Outcome

How far a walk got, and why it stopped.

Constants

MAX_INDEXED

The most files one walk keeps. The time budget bounds a walk first in practice; this bounds its memory on a tree fast enough to list a million names inside it.

Functions

score

Score query against the files in index, keeping the best limit to list.

search_root

Where a search should start, given where the user is.

walk

Walk root for datasets, handing batches to emit as they are found.

walk_up_to

walk, keeping at most cap files.

walk_with_specs

walk, keeping the files formats reads as well, as the listing names them: by a spec’s glob, or by its magic in the first bytes of a file whose name says nothing, at most [crate::discover::MAX_SNIFFS_PER_DIR] of them a directory.