8% HTML without a single html file in the repo

[shaderecker]

Hi,
my repo https://github.com/shaderecker/ansible-pihole shows 8% HTML language, but there are no html files in there.
When clicking on the language tag, it shows no results (since there are no html files).

Do you have any ideas why HTML is being detected here?

This error only happens directly on GitHub.
Running linguist locally works as expected and doesn't show the wrong HTML statistics.

$ github-linguist . --breakdown
94.35%  YAML
5.65%   Shell

YAML:
bootstrap-pihole.yaml
inventory.yaml
keepalived.yaml
roles/bootstrap/tasks/main.yaml
roles/docker/tasks/main.yaml
roles/keepalived/tasks/main.yaml
roles/pihole/tasks/main.yaml
roles/sshd/tasks/main.yaml
roles/start_keepalived/tasks/main.yaml
roles/stop_keepalived/tasks/main.yaml
roles/sync/tasks/main.yaml
roles/updates/tasks/main.yaml
sync.yaml
update-pihole.yaml

Shell:
roles/keepalived/files/check_pihole.sh
roles/sync/templates/pihole_sync.j2

[lildude]

🤔 Looks like you might be running an old version of Linguist. This is what I see, which matches what your repo reports on GitHub:

$ github-linguist ansible-pihole --breakdown
86.68%  YAML
7.37%   HTML
5.95%   Shell

YAML:
bootstrap-pihole.yaml
inventory.yaml
keepalived.yaml
roles/bootstrap/tasks/main.yaml
roles/docker/tasks/main.yaml
roles/keepalived/tasks/main.yaml
roles/pihole/tasks/main.yaml
roles/sshd/tasks/main.yaml
roles/start_keepalived/tasks/main.yaml
roles/stop_keepalived/tasks/main.yaml
roles/sync/tasks/main.yaml
roles/updates/tasks/main.yaml
sync.yaml
update-pihole.yaml

Shell:
roles/keepalived/files/check_pihole.sh
roles/sync/templates/pihole_sync.j2

HTML:
roles/keepalived/templates/keepalived.j2
$

.j2 is associated with "HTML+Django" which is grouped under HTML:

https://github.com/github/linguist/blob/7e21cd427be5ab1bff72ed6e3d3834474145ea88/lib/linguist/languages.yml#L2097-L2113

Now comes the interesting question: "Why is .j2 being detected as Shell and HTML?

Because the roles/sync/templates/pihole_sync.j2 file has told us it's a shell script in the very first line and this takes precedence over the extension.

[Alhadis]

On an unrelated side-note, your shell-script should begin with

#!/usr/bin/env bash

instead of

#!/bin/bash

if you need your script to be specifically run by Bash, as opposed to /bin/sh, which generally maps to the system's default shell. The reason is that not every Unix-like system has /bin/bash installed; e.g., many BSD derivatives use ksh(1) or csh(1), and a future release of macOS will remove the system-installed Bash once Apple have finished migrating users to Zsh.

</diatribe>

[shaderecker]

Thanks for the explanation @lildude
And I was indeed running an older version (7.9.0) from Ubuntu 20.10 sources.

One question: Why does it then display no results when I click on the HTML lang tag (https://github.com/shaderecker/ansible-pihole/search?l=html)?
The .j2 file is categorized in https://github.com/shaderecker/ansible-pihole/search?l=shell (due to the shebang)
Wouldn't it be better if it appeared in both categories then? It really confused me that there are no results for HTML.

And I also have another .j2 file (roles/keepalived/templates/keepalived.j2) that I would then expect to appear in the HTML search.

And @Alhadis thanks for the hint about the shebang, I will keep it in mind.

[lildude]

One question: Why does it then display no results when I click on the HTML lang tag (https://github.com/shaderecker/ansible-pihole/search?l=html)?
...
It really confused me that there are no results for HTML.
...
And I also have another .j2 file (roles/keepalived/templates/keepalived.j2) that I would then expect to appear in the HTML search.

That's because Search doesn't know about the language grouping. Linguist has said this is "HTML" (that's where the html comes from in the query) so that's what Search returns.

The .j2 file is categorized in https://github.com/shaderecker/ansible-pihole/search?l=shell (due to the shebang)
Wouldn't it be better if it appeared in both categories then?

No. You've explicitly told Linguist that this is Shell and not HTML (or anything else) by using the shebang. Linguist will not go against this choice. Linguist can also only associate a file with one language at a time.

[shaderecker]

That's because Search doesn't know about the language grouping. Linguist has said this is "HTML" (that's where the html comes from in the query) so that's what Search returns.

But the html search doesn't return any results. That part I don't get. 🙂

[lildude]

Whoops. Ignore what was here before. It's because the actual language is "HTML+Django" and not HTML, but Linguist is returning the group, not the language, so search is using that.

[shaderecker]

Aha ok :D Now I understand the behaviour on GitHub.
Thank you for your answers! 😃