Extreme amount of incorrect language detected

[gfcwfzkm]

For some odd reason, the language on my TinyTapeout Repository https://github.com/gfcwfzkm/tt09-scope-bfh-mht1_3 is detected heavily as F# - despite containing no code with F#. It is so extreme, that my whole githup account statistics show now F# as the dominant language.

On top of that, in the same repo, a considerable amount of Verilog is detected, yet when searched for it, only 6 little files contain Verilog at all.

Any idea how I can fix that?

[lildude]

For some odd reason, the language on my TinyTapeout Repository https://github.com/gfcwfzkm/tt09-scope-bfh-mht1_3 is detected heavily as F# - despite containing no code with F#. It is so extreme, that my whole githup account statistics show now F# as the dominant language.

The reason you are seeing this is because of these two files:

F#:
fpga/gowin_vhdl_tang25k/impl/pnr/ao_0.fs
fpga/gowin_vhdl_tang25k/impl/pnr/tt_project_env.fs

The .fs extension is associated with multiple languages so we rely on this heuristic to try and determine the language:

linguist/lib/linguist/heuristics.yml

Lines 277 to 286 in f164d13

	- extensions: ['.fs']
	rules:
	- language: Forth
	pattern: '^(: |new-device)'
	- language: 'F#'
	pattern: '^\s*(#light|import|let|module|namespace|open|type)'
	- language: GLSL
	pattern: '^\s*(#version|precision|uniform|varying|vec[234])'
	- language: Filterscript
	pattern: '#include|#pragma\s+(rs|version)|__attribute__'

If this doesn't match, things fall through to the classifier which makes a best effort guess based on the samples we have, which in this case isn't the right guess.

Both of these files are pretty large at 5.7MB each so nearly 12MB of your entire repo is considered F#. These won't appear in the search results because of their size (the number in the second column is in bytes):

$ github-linguist .
93.21%  11948739   F#
3.83%   490822     HTML
1.75%   223885     Verilog
1.18%   151042     VHDL
0.01%   1216       Tcl
0.01%   1103       Makefile
0.01%   1020       Dockerfile
0.01%   755        Stata
0.01%   679        Python
0.00%   106        Shell
$

On top of that, in the same repo, a considerable amount of Verilog is detected, yet when searched for it, only 6 little files contain Verilog at all.

This would be due to these file:

Verilog:
fpga/gowin_vhdl_tang25k/impl/gao/ao_0/gw_ao_expression.v
fpga/gowin_vhdl_tang25k/impl/gao/ao_0/gw_ao_init.v
fpga/gowin_vhdl_tang25k/impl/gao/ao_0/gw_ao_parameter.v
fpga/gowin_vhdl_tang25k/impl/gao/ao_0/gw_ao_top_define.v
fpga/gowin_vhdl_tang25k/impl/gao/ao_control/gw_con_parameter.v
fpga/gowin_vhdl_tang25k/impl/gao/ao_control/gw_con_top_define.v
fpga/gowin_vhdl_tang25k/impl/gao/gao.v
fpga/gowin_vhdl_tang25k/impl/gao/gw_gao_top.v
test/dummy.v
test/tb.v

... with fpga/gowin_vhdl_tang25k/impl/gao/gao.v being 209kb which is most of the 223885b shown. Some missing files in the search results are likely due to indexing and caching, which Linguist has no control over. They may appear in future.

Any idea how I can fix that?

You can use an override.

[gfcwfzkm]

Fascinating, thank you!
I was searching all over because I thought that the tool would actually look into the source files. The .fs files pretty much just contains 0 and 1s.

I will use the override argument as provided. Thank you a lot!