Tools used in creation of the Granite keyboard layout. Main features:
- Toolkit for scoring cost for key sequences ("ngrams"): [Docs]
- Ngram frequency analysis tools:
ngram_showandngram_compare. [Docs]
You can use uv and let it handle python (virtual) environments for you. In this case
- install uv
- Clone this repo or just download the contents
- Use
uv run <command>instead of<command>(must be executed in this directory). It's also possible to activate the virtual environment in the created.venvdirectory (after first run), if you want that. - This might be the right option for you if you are not familiar with python and virtual environments and it's a bit more hassle free.
- Use python 3.12 in a fresh virtual environment (this is important as package versions are pinned so you may mess up your system if you don't)
python -m pip install git+https://github.com/fohrloop/granite-tools.git
The following assumes that the ngram files follow the same format as the Keyboard Layout Optimizer; <frequency> <chars> on each line. For example:
7.851906052482727
7.534280931977296 e
5.621867629166546 t
4.477877757676168 r
4.430916637427879 a
and that each set of ngrams are called 1-grams.txt, 2-grams.txt and 3-grams.txt within a single folder, like this:
📁 ngrams/
├─📁 leipzig/
| ├─📄 1-grams.txt
| ├─📄 2-grams.txt
| └─📄 3-grams.txt
└─📁 tldr17/
├─📄 1-grams.txt
├─📄 2-grams.txt
└─📄 3-grams.txt
❯ ngram_show ./ngrams/kla-english -s 2 --plot -n 10
────────────────── kla-english ───────────────────
1: e␣ ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 2.88
2: ␣t ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 2.31
3: th ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1.98
4: he ▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 1.88
5: s␣ ▇▇▇▇▇▇▇▇▇▇▇▇▇ 1.75
6: ␣a ▇▇▇▇▇▇▇▇▇▇▇▇▇ 1.73
7: d␣ ▇▇▇▇▇▇▇▇▇▇▇▇ 1.52
8: in ▇▇▇▇▇▇▇▇▇▇▇ 1.46
9: t␣ ▇▇▇▇▇▇▇▇▇▇ 1.32
10: er ▇▇▇▇▇▇▇▇▇▇ 1.28
Compare two ngram corpora. Example:
❯ ngram_compare /home/fohrloop/code/granite-english-ngrams/ngrams/english/ /home/fohrloop/code/granite-code-ngrams/ngrams/code/ --plot -s 3 -n 20 -
i --diff -w
─────────────────────english────────────────────── ───────────────────────code───────────────────────
1: the ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 2.49 1 (+2826): --- ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 0.61
2: ing ▇▇▇▇▇▇▇▇▇▇▇ 1.43 2 ( +4): ion ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 0.47
3: and ▇▇▇▇▇▇▇▇▇▇ 1.26 3 ( +6): ent ▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 0.42
4: hat ▇▇▇▇▇ 0.61 4 ( +33): con ▇▇▇▇▇▇▇▇▇▇▇▇▇ 0.41
5: her ▇▇▇▇▇ 0.60 5 ( +7): tio ▇▇▇▇▇▇▇▇▇▇▇▇▇ 0.39
6: ion ▇▇▇▇▇ 0.56 6 ( -5): the ▇▇▇▇▇▇▇▇▇▇ 0.31
7: tha ▇▇▇▇ 0.55 7 ( -5): ing ▇▇▇▇▇▇▇▇▇▇ 0.30
8: for ▇▇▇▇ 0.55 8 ( +17): ate ▇▇▇▇▇▇▇▇▇▇ 0.30
9: ent ▇▇▇▇ 0.53 9 ( +275): sel ▇▇▇▇▇▇▇▇▇▇ 0.29
10: thi ▇▇▇▇ 0.50 10 ( +194): ass ▇▇▇▇▇▇▇▇▇ 0.27
11: all ▇▇▇▇ 0.47 11 ( +68): ect ▇▇▇▇▇▇▇▇▇ 0.26
12: tio ▇▇▇▇ 0.45 12 ( +37): ons ▇▇▇▇▇▇▇▇ 0.25
13: ver ▇▇▇ 0.42 13 ( +88): ort ▇▇▇▇▇▇▇▇ 0.25
14: you ▇▇▇ 0.42 14 ( +240): ser ▇▇▇▇▇▇▇▇ 0.25
15: ter ▇▇▇ 0.40 15 ( +456): elf ▇▇▇▇▇▇▇▇ 0.24
16: ere ▇▇▇ 0.38 16 ( +890): def ▇▇▇▇▇▇▇▇ 0.24
17: his ▇▇▇ 0.38 17 ( +64): ame ▇▇▇▇▇▇▇▇ 0.23
18: ith ▇▇▇ 0.36 18 ( +214): por ▇▇▇▇▇▇▇▇ 0.23
19: wit ▇▇▇ 0.35 19 ( -4): ter ▇▇▇▇▇▇▇ 0.23
20: was ▇▇▇ 0.33 20 ( +30): est ▇▇▇▇▇▇▇ 0.22
25: ate ▇▇▇ 0.32 22 ( -14): for ▇▇▇▇▇▇▇ 0.22
37: con ▇▇ 0.25 60 ( -47): ver ▇▇▇▇▇ 0.16
49: ons ▇▇ 0.22 71 ( -68): and ▇▇▇▇ 0.13
50: est ▇▇ 0.22 95 ( -84): all ▇▇▇▇ 0.11
79: ect ▇ 0.17 111 ( -101): thi ▇▇▇ 0.10
81: ame ▇ 0.16 127 ( -110): his ▇▇▇ 0.10
101: ort ▇ 0.15 141 ( -136): her ▇▇▇ 0.09
204: ass ▇ 0.10 145 ( -127): ith ▇▇▇ 0.09
232: por ▇ 0.09 155 ( -136): wit ▇▇▇ 0.09
254: ser ▇ 0.08 189 ( -173): ere ▇▇ 0.07
284: sel ▇ 0.08 505 ( -498): tha ▇ 0.04
471: elf 0.05 603 ( -599): hat ▇ 0.03
906: def 0.03 1022 (-1008): you ▇ 0.02
2827: --- 0.00 3196 (-3176): was 0.01