Incorrect language shown for repository

[dEhiN]

🏷️ Discussion Type

Question

Body

I'm not sure if this is a bug or not, but I noticed that for one of my repositories, an incorrect language seems to be shown in the stats. Specifically, my repo CompStart shows the following as the language breakdown for the project:

However, when I click on HTML, there are no code files. This makes sense since the project is all Python, PowerShell, and Batch. The HTML content seems to be from 55 issues and 3 pull requests. Here's a screenshot:

In some of my Markdown files, I have used HTML elements, notably <br>, and I may have done similarly in issue or PR comments. But, should these count? The language breakdown makes it seem like most of the project is HTML, when it has nothing to do with HTML.

So, I don't know if this is a bug or if there's a way to not have issues and PRs included in the language breakdown. Any ideas?

Thanks in advance!

[Gurnoor2910]

1. Force the correct main language

In your repo root, create .gitattributes and add one line like this (adjust extension/language):

*.js linguist-language=JavaScript

or

*.py linguist-language=Python

Then:

• git add .gitattributes
• git commit -m "fix language"
• git push

GitHub will recompute the language in a few minutes.

---

2. Hide misleading files from stats

If HTML / CSS / logs are dominating, ignore them:

*.html linguist-detectable=false
*.css linguist-detectable=false
*.log linguist-detectable=false

Commit and push again; the language bar will update.

---

3. If it still looks wrong

• Make sure the real code files (e.g., .js, .py) are the majority by lines of code.
• Delete or move huge non‑code files (dumps, logs) that skew the stats.

[dEhiN]

Hi,

Thank you for your reply! I shared below that there is a Python module I use called PyInstaller that builds exe files from Python scripts. It does this by bundling the runtime environment. In order to do this, it creates a lot of intermediary files, and one of them seems to be a pretty large HTML file.

I didn't realize this and through my testing of the tool, I have multiple of these files generated. I'm going to create the .gitattribute file as you all have mentioned.

I do have a question though: If I did force the correct main language, would the other languages still be detected? I suppose it would just change the percentages?

[rushikesh-bobade]

Hello!

To answer your specific question: No, GitHub does not count Issues, Pull Requests, or Comments in its language statistics. The tool GitHub uses (called Linguist) only calculates stats based on the actual files committed to your repository's default branch.

If HTML is showing up as the dominant language, it isn't coming from
tags in your markdown or comments. It is highly likely that there is a large generated HTML file, a docs folder, or some vendor/library files committed to your repo that you might have overlooked. (Linguist calculates by file size/bytes, so even one large raw HTML file can easily outweigh dozens of Python scripts).

To fix this, you don't need to change your code. You just need to tell Linguist to ignore those specific files so your Python and PowerShell stats shine through.

How to fix it:

1. Create a file named .gitattributes in the root of your repository.

2. To completely hide all HTML from the stats, add this exact line to the file:
   *.html linguist-detectable=false

3. Commit and push the .gitattributes file.

Note: It might take a few minutes or a browser refresh for the language bar to recalculate and update. Let me know if this fixes it for you!

[dEhiN]

Hi,

Thank you for your reply! I shared below that there is a Python module I use called PyInstaller that builds exe files from Python scripts. It does this by bundling the runtime environment. In order to do this, it creates a lot of intermediary files, and one of them seems to be a pretty large HTML file.

I didn't realize this and through my testing of the tool, I have multiple of these files generated. I'm going to create the .gitattribute file as you all have mentioned.

[Gecko51]

Issues, PRs, and markdown content don't count toward language stats. Linguist only looks at actual committed files in the repo.

So if HTML is showing up with a significant percentage, there's a real file in your repo that Linguist is picking up. Could be something you forgot was there: a generated HTML report, a docs folder, or a large file that got committed at some point. You can find it quickly with:

git ls-files | grep -i "\.html"

Once you know what it is, add it to .gitattributes in your repo root:

*.html linguist-detectable=false

Or if it's a whole directory (like a docs/ or dist/ folder):

docs/ linguist-vendored=true

linguist-vendored removes the path from stats entirely, useful for generated or third-party content. linguist-detectable=false just hides it from the language bar but keeps it tracked.

Push the .gitattributes file and the language bar updates within a few minutes after the next push.

[dEhiN]

Hi,

Thank you for the suggested command to run! So, I'm using PyInstaller, a Python module that builds exe files from Python scripts by bundling in the runtime environment. It appears that as part of the process, the tool creates an HTML file that's quite large. Additionally, through testing and whatnot, I've run the tool several times, so each time, an HTML file was generated.

Do you know why the search didn't show any code files but only issues and PRs?

Anyway, thank you once again!