[Question] Extract labels from PR title

[sebastianrosch]

I have tried lots of different configurations but can't figure out how to extract the label from the PR title. No matter what I do, I always get only uncategorized PRs.

The latest config I'm using looks like this:

 - name: Generate release changelog
        id: generate_release_changelog
        uses: mikepenz/release-changelog-builder-action@v4
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        with:
          configurationJson: |
            {
              "label_extractor": [
                {
                  "pattern": ".*(\\[Feature\\]|\\[Fix\\]|\\[Chore\\]|\\[Test\\]).*",
                  "on_property": "title",
                  "target": "$1"
                }
              ]
            }

I've looked at similar issues and found a reference to the test here: https://github.com/mikepenz/release-changelog-builder-action/blob/dffddcd83353ebb1c64f53b6cacfd6e24b707728/__tests__/transform.test.ts#L145

However, that's not working for me either.

Can you point out what I'm doing wrong?

[mikepenz]

Good day.

So before looking into the extractor itself, I'd like to point out that in your case you use the default configuration for categories, so chore would not be recognized as a category. The default categories are: https://github.com/mikepenz/release-changelog-builder-action/blob/v4.2.2/src/configuration.ts#L76-L93

That said you can probably use something like this instead: https://github.com/mikepenz/release-changelog-builder-action/issues/1172#issuecomment-1650493613 whichever is first in [] to be the label. that also scales better for the future.

Unless you already used that too?

[sebastianrosch]

Hi @mikepenz, thanks for the quick response.

I tried lots of different configurations and also defined them explicitly.

For example:

{
  "template": "#{{CHANGELOG}}\n\n<details>\n<summary>Uncategorized</summary>\n\n#{{UNCATEGORIZED}}\n</details>",
  "categories": [
    {
      "title": "## 🚀 Features",
      "labels": [
        "feat",
        "feature"
      ]
    },
    {
      "title": "## 🐛 Fixes",
      "labels": [
        "fix",
        "bug"
      ]
    },
    {
      "title": "## 🛠️ Chores",
      "labels": [
        "chore"
      ]
    },
    {
      "title": "## 🛠️ Dependencies",
      "labels": [
        "deps"
      ]
    }
  ],
  "label_extractor": [
    {
      "method": "match",
      "pattern": "^\\[(\\w*)\\]",
      "target": "$1",
      "on_property": "title"
    }
  ]
}

In my understanding of the regex, it should match Chore in the case of a PR titled [Chore] ..... I also tried defining the labels in the categories as title case. I've played around with the regex in regex tools to see if that works, then escaped the \. I still get all PRs as uncategorized. It seems like it reads the JSON correctly as the template looks different.

Am I not getting something regarding the escaping of the regex or the definition of the labels?

Thanks!

[mikepenz]

Is the repo you run those on open source? Would allow me to reproduce it more quickly

[sebastianrosch]

The repos I'm trying to do this for are not, but I was able to reproduce it in this public repo: https://github.com/sebastianrosch/changelog-test

[mikepenz]

That is super helpful. Using your demo repo. Here's an adjusted config which should handle your use-case:

{
    "template": "#Changes: #{{CHANGELOG}}\n\n<details>\n<summary>Uncategorized</summary>\n\n#{{UNCATEGORIZED}}\n</details>",
    "categories": [
        {
            "title": "## 🚀 Features",
            "labels": [
                "[feat]",
                "[feature]",
                "[Feature"
            ]
        },
        {
            "title": "## 🐛 Fixes",
            "labels": [
                "[fix]",
                "[bug]",
                "[Fix]"
            ]
        },
        {
            "title": "## 🛠️ Chores",
            "labels": [
                "[chore]",
                "[Chore]"
            ]
        },
        {
            "title": "## 🛠️ Dependencies",
            "labels": [
                "[deps]"
            ]
        }
    ],
    "label_extractor": [
        {
            "method": "match",
            "pattern": "\\[.+?\\]",
            "on_property": "title"
        }
    ]
}

Key details:

• fixed the template, the # was missing in front of CHANGELOG
• adjusted the regex to match the [] bracket. note this will match including the bracket
• adjusted the labels in the categories to include the brackets as well

That said, v5 of the action actually contains a feature that makes such regex's much simpler, by offering a regex match logic similar to what regexr does.

So for @v5 the following will work:

{
    "template": "#Changes: #{{CHANGELOG}}\n\n<details>\n<summary>Uncategorized</summary>\n\n#{{UNCATEGORIZED}}\n</details>",
    "categories": [
        {
            "title": "## 🚀 Features",
            "labels": [
                "feat",
                "feature",
                "Feature"
            ]
        },
        {
            "title": "## 🐛 Fixes",
            "labels": [
                "fix",
                "bug",
                "Fix"
            ]
        },
        {
            "title": "## 🛠️ Chores",
            "labels": [
                "chore",
                "Chore"
            ]
        },
        {
            "title": "## 🛠️ Dependencies",
            "labels": [
                "deps"
            ]
        }
    ],
    "label_extractor": [
        {
            "method": "regexr",
            "pattern": "\\[(.+?)\\].+",
            "target": "$1",
            "on_property": "title"
        }
    ]
}

[sebastianrosch]

Awesome, thank you, that worked! I've tried the regexr method before but didn't see that this was only supported in v5. Thanks for the help!