Better position information

[Alexey Alekhin]

Right now all errors are reposted with position 1:1. As I see from the travis lint output, it gives you information about the location in terms of the YAML structure. So I guess with some simple YAML parsing this could be translated to a line:column position for the linter..

[Matthew Schuchard]

I don't see this kind of information in 1.8.8. Could you show an example?

[Alexey Alekhin]

Here's an example config:

sudo: false
language: scala
jdk: oraclejdk8
scala: 2.12.3
script: sbt test
notifications:
  webhooks:
    urls: https://webhooks.gitter.im/e/foo
    on_start: false           # <-- problem here
cache:
  directorie:                 # <-- problem here
    - $HOME/.ivy2/cache
    - $HOME/.sbt/boot/

And here is the linter's output:

>  travis lint --no-interactive .travis.yml
Warnings for .travis.yml:
[x] in notifications.webhooks section: dropping on_start section: illegal value false
[x] value for cache section is empty, dropping
[x] in cache section: unexpected key directorie, dropping

As you every warning has some clue about location:

• in notifications.webhooks section, on_start section
• value for cache section
• in cache section

[Matthew Schuchard]

So typically the way we do these Atom linter packages is to parse the linter output information for line/col information and then convert it back into an int for display.

What you are proposing is:

• Regexp out the section name.
• Use a YAML parser to grab the line information for the key.
• Display the line information.

1 and 3 are certainly possible. I have not done YAML parsing in node.js before, so I do not know what kind of support the YAML module has. If I can return line info for a key in a file, then it is certainly possible.

Thanks for the suggestion. Let me know if any of this sounds off from what you are suggesting. I will look into this feature.

[Alexey Alekhin]

@mschuchard you completely understood my point. I realise that this is probably not an easy feature to request, but just wanted to submit this idea. Also I'm pretty sure there are plenty npm packages for parsing YAML..

I found out that somebody already tried to request better linter output for the travis CLI, but it didn't result in anything: travis-ci/travis.rb#203. Although the idea of querying API directly is probably worth further investigating.

[Alexey Alekhin]

By calling the API directly you can at least eliminate step 1 (matching sections names in the linter output). Calling

curl -T .travis.yml api-staging.travis-ci.org/lint

on the example config from above returns this JSON:

{
  "lint": {
    "warnings": [
      {
        "key": [
          "notifications",
          "webhooks"
        ],
        "message": "dropping \"on_start\" section: illegal value \"false\""
      },
      {
        "key": [],
        "message": "value for \"cache\" section is empty, dropping"
      },
      {
        "key": [
          "cache"
        ],
        "message": "unexpected key \"directorie\", dropping"
      }
    ]
  }
}

So key array gives you the path in the YAML structure.