Metadata-Version: 2.1
Name: searchf
Version: 1.2.1
Summary: A text-based application running in terminals to search into text files
Home-page: https://github.com/human3/searchf
Author: Adrien Felon
License: Apache 2.0
Keywords: search,file,curses
Platform: UNKNOWN
Description-Content-Type: text/markdown
License-File: LICENSE

# [searchf](https://github.com/human3/searchf)

Utility program to interactively search and colorize keywords in text files, select relevant lines, remove noise and help you discover the things you should care about. Works in terminals and uses stack-based interactions to minimize the number of key presses (think HP calculators).

![Screenshot searchf 1](https://user-images.githubusercontent.com/15265841/147424844-9bece2d4-ceb0-4ea1-b989-a40ea3e6d3ac.png)

The above screenshot shows how the content of a "random" text file (`app.py` being the source of this utility) can be filtered using 3 filters:
- `def AND init` reveals 6 lines in red
- `def AND push` reveals 3 lines in orange
- `keyword` reveals 40 lines in yellow

Keywords can be added to the most recent filter (the one at the bottom and associated to yellow) by pressing `+` or `=`, and removed/poped from the filter by pressing `-`. New filters can be pushed to the stack by pressing `ENTER` or `f`, and poped by pressing `backspace` or `delete`.

A filter is a list of keywords that a line must contain to match and get highlighted in a specific color. All keywords in the same filter are ANDed together. By defining multiple filters, you can reveal more content of the file (filters are ORed...). By default, lines not matching any filter are hidden, but their visibility can be toggled by pressing `m`. Filters are evaluated in the order that they are defined, meaning lines are shown in the color of the first filter they match.

## Features

- Supports multiple views (try pressing `1`, `2`, `3`)
- Supports multiple palettes (try pressing `c` for color)
- Various display modes (`l` toggle line numbers, `m` toggle non-matching line, ...)
- `less` like search key bindings (`/`, then `n` for next, `p` for previous)

## Installation

`pip install searchf`

then run

`searchf-test`

to run builtin application tests and verify your installation. Please note that your terminal must supports color (`TERM=screen-256color` or `TERM=xterm-256color`) and must be UTF-8 enabled (eg. start tmux with "tmux -u").

## Usage

`searchf <FILE>`

- Press `f` to enter keyword in a new filter
- Press `?` for help
 
![Screenshot searchf help](https://user-images.githubusercontent.com/15265841/147424944-cbb41951-9911-4577-bd3a-857293802f0a.png)

When all lines are shown, including the ones not matching any filter, it can be usefull to scroll to next match with `n` or previous one with `p`. When all lines are displayed, it can be hard to identify which ones that are matching which filter, so you can either show the line numbers by pressing `l`, which will show the matching lines with colorized output, or change the highlight mode by pressing `h`, to colorize lines as a whole (and not just keywords). Just try it, as it's likely more understandable by doing than by reading this...

Showing all lines, with wrapping and numbers enabled, colorizing lines as a whole instead of just keywords:

![Screenshot searchf 2](https://user-images.githubusercontent.com/15265841/147425069-609e346d-c84d-452c-bfb2-8e32cadf10d5.png)

## Why this utility?

This tool is born from my need to be able to efficiently explore log files in interactive fashion, searching for cues, hitting dead ends and backtracking (hence the use of push/pop of filter and keyword), but also going down some exploratory paths on the side (hence the support of views).

To be more specific, I had to dig into build log files, which were a raw aggregate of many heterogenous sources (numerous compiler output, deployment scripts, test run and results, ...) resulting in rather unstructured output. The only commonality being that everything was somehow and more or less line-oriented... When a build failure occured, hints of the root caused could be hiding about anywhere.

So this tool ended making little or no assumption on the input file, which can be anything, unstructured and heterogenous, as long as it is line oriented.

## Development

If working from the sources (ie not an installed package), the application can be launched as a module:

`python3 -m searchf.app <FILE>`

To run all tests:

`python3 -m searchf.test.all`

Some unit tests can get triggerred through `pytest`.

To get coverage report (requires coverage package):

`coverage run -m searchf.test.all`, then
`coverage html`

As of version 1.2:

![Screenshot searchf coverage](https://user-images.githubusercontent.com/15265841/147427412-9ac304b6-c0d1-40fe-bc8a-b8539af7f5c4.png)

## Known Issues

- Does not work on Windows... No color there... It can run, but it's hard:
  - py -m pip install windows-curses
    https://www.lfd.uci.edu/~gohlke/pythonlibs/#curses
  - py -m pip install "YOUR\_DOWNLOADS\curses-2.2.1+utf8-cp310-cp310-win_amd64.whl"
  - mingw64 + winpty py searchf.py

## Wish list

- More palettes (and move them into separate module)
- Ability to save/load filters
- Ability to pass current filters to another view
- Tail function that automatically reload (instead of having to press `t`)
- Screen dynamic resizing
- Bookmarks
- Ability to save filtered file
- Cleanup: remove COLOR_BAR hack in source (inner beauty's sake)


