Monadoc
The latest version of talash is 0.3.0-0.

talash

Version 0.1.1.0 revision 0 uploaded by rahguzar.

Package meta

Synopsis
Line oriented fast enough text search
Description

This library provides searching a large number of candidates against a query using a given style. Two styles are provided. The default is orderless style in which a match occurs if the words in the query all occur in the candidate regardless of the order of their occurrence. A fuzzy style is also provided in which a candidate matches if all the characters of the query occur in it in order.

There is also a TUI searcher/selector interface provided using a brick app. Like an extremely barebones version of fzf and mostly intended to be a starting point that has to be configured according to the needs or else it can be embedded into other applications to provide a selection interface.

There is also a piped searcher/seeker provided in which searcher runs in the background and can be used by a seeker communicating with it using named pipes.

The is also a demo executable for both the brick app and piped version that gets the candidates for the stdin. Use talash help for usage information.

Some care has been taken to make the searcher performant. On my laptop searching using the tui app to search all files in my /usr/local with 60K items, the search results appear almost instantly. Searching among about 340K files in /usr/share there is some but bearable lag between the keypresses and the search results. While searching between 1264K files, that fd finds from / there is a lag of a second or so before the results appear. The three scenarios consume about 50 MiB , 130 MiB and 500 MiB of memory respectively which is almost entirely due to the Vector Text storing the candidates.

The nice string matching interface provided by alfred-margaret is responsible for a big part of the performance. While vector-sized is responsible for most of memory efficieny. Performance can potentially be further improved by using all the cores but it is good enough for my typical use cases of searching among a few thousand or at most a few tens of thousands of candidates. As a result parallel matching is unlikely to be implemented.

The package is lightly maintained, bugs reports are welcome but any action on them will be slow. Patches are welcome for 1. bugfixes 2. simple performance improvements 3. Adding mouse bindings to tui 4. New search styles, especially a better fuzzy one, that matches each word in the query fuzzily but the words themselves can be matched in any order (I am not sure what is a sensible implementation of this).

Author
Rahguzar
Bug reports
n/a
Category
search, tui
Copyright
n/a
Homepage
n/a
Maintainer
aikrahguzar@gmail.com
Package URL
n/a
Stability
n/a

Components