#unicode #word #count #wc

app uwc

Counts things in unicode text files

2 unstable releases

✓ Uses Rust 2018 edition

0.2.0 Jan 29, 2019
0.1.0 Feb 5, 2018

#37 in Text processing

17 downloads per month

MPL-2.0 license

33KB
622 lines

uwc

Build Status crates.io page

Like wc, but unicode-aware, and with line mode.

uwc can count:

  • Lines
  • Words
  • Bytes
  • Grapheme clusters
  • Unicode code points

Additionally, it can operate in line mode, which will count things within lines.

Usage example

By default, uwc will count lines, words, and bytes. You can specify the counters you'd like, or ask for all counters with the -a flag.

$ uwc tests/fixtures/**/input
lines  words  bytes  filename
8      5      29     tests/fixtures/all_newlines/input
0      0      0      tests/fixtures/empty/input
0      0      0      tests/fixtures/empty_line_mode/input
1      9      97     tests/fixtures/flags_bp/input
1      9      97     tests/fixtures/flags_cl/input
1      9      97     tests/fixtures/flags_w/input
0      1      5      tests/fixtures/hello/input
1      9      97     tests/fixtures/i_can_eat_glass/input
8      8      29     tests/fixtures/line_mode/input
7      8      28     tests/fixtures/line_mode_no_trailing_newline/input
7      8      28     tests/fixtures/line_mode_no_trailing_newline_count_newlines/input
34     66     507    total

$ uwc -a tests/fixtures/**/input
lines  words  bytes  graphemes  codepoints  filename
8      5      29     23         24          tests/fixtures/all_newlines/input
0      0      0      0          0           tests/fixtures/empty/input
0      0      0      0          0           tests/fixtures/empty_line_mode/input
1      9      97     51         51          tests/fixtures/flags_bp/input
1      9      97     51         51          tests/fixtures/flags_cl/input
1      9      97     51         51          tests/fixtures/flags_w/input
0      1      5      5          5           tests/fixtures/hello/input
1      9      97     51         51          tests/fixtures/i_can_eat_glass/input
8      8      29     28         28          tests/fixtures/line_mode/input
7      8      28     27         27          tests/fixtures/line_mode_no_trailing_newline/input
7      8      28     27         27          tests/fixtures/line_mode_no_trailing_newline_count_newlines/input
34     66     507    314        315         total

You can also switch into line mode with the --mode flag:

$ uwc -a --mode line tests/fixtures/line_mode/input
lines  words  bytes  graphemes  codepoints  filename
0      1      1      1          1           tests/fixtures/line_mode/input:1
0      1      2      2          2           tests/fixtures/line_mode/input:2
0      1      3      3          3           tests/fixtures/line_mode/input:3
0      1      5      4          4           tests/fixtures/line_mode/input:4
0      1      1      1          1           tests/fixtures/line_mode/input:5
0      1      4      4          4           tests/fixtures/line_mode/input:6
0      1      2      2          2           tests/fixtures/line_mode/input:7
0      1      3      3          3           tests/fixtures/line_mode/input:8
0      8      21     20         20          tests/fixtures/line_mode/input:total

Why?

The goal of this project is to consider unicode rules correctly when counting things. Specifically, it should:

  • Count all newline characters correctly. This includes lesser-known line breaks, like NEL (U+0085), FF (U+000C), LS (U+2028), and PS (U+2029).
  • Count all words using the Unicode standard's word boundary rules.
  • Count all complete grapheme clusters correctly, so that even edge cases like Z҉͈͓͈͎a̘͈̠̭l̨̯g̶̬͇̭o̝̹̗͎̙ ͟t͖̙̟̹͇̥̝͡e̥͘x͚̺̭̻͘t͉͔̩̲̘, for example, are counted correctly.

It does not aim to implement these unicode algorithms, however, so it makes use of the unicode-segmentation library for most of the heavy lifting. And since Unicode support in the Rust ecosystem is not quite mature yet, that has some consequences for this project. See the caveats below.

It is primarily a fun side project for me, and an excuse to learn more about Rust and unicode.

Installation

It is published on crates.io, so simply:

$ cargo install uwc

Caveats

UTF-8

It only supports UTF-8 files. UTF-16 can go on my to-do list if there is demand. For now, you can use iconv to convert non-UTF-8 files first.

Speed

It is slower than wc. My analysis hasn't been extensive, but as far as I can tell, the reasons are:

  • It is using unicode algorithms, which are just going to be slower than ASCII no matter what.
  • I am not that experienced with Rust, so it's quite possible I'm not doing something as efficiently as possible.
  • My free time is limited, and I am prioritizing correctness over speed (though speed is good).

With that said, parallelization helps. With testing on my local laptop with larger data sets, the speed is within an order of magnitude of wc. I measured uwc being 3x slower than wc on a collection of 18 MiB of text files.

Localization

Rust, as yet, has no localization libraries, so this has some consequences. Some counts will just be wrong, such as hyphenated words, which is locale-specific and requires language dictionary lookups to be correct. Also, there are some languages that have no syntactic word separators, such as Japanese, so e.g.

ガラス食べられます

should be 5 words, but without localization, we cannot determine that.

Dependencies

~6.5MB
~108K SLoC