[25.1] Avoid missing field value failure in data tables

[guerler]

Resolves #21183. Exploring options to resolve a failure caused by injecting an empty name column into the data table. Because only the column name was added without corresponding field data, the column list becomes longer than the field list. As a result, when searching for a column such as path, the returned index is shifted by one, leading to an out-of-range error when accessing the corresponding field value. twobit data tables do not include a name field, and since the path field is the last column, any attempt to access it for downloading data from IGV results in a failure. In my view, the correct approach is to avoid injecting an empty name column, since a value column is already required and guaranteed to exist. Consumers of this API endpoint should instead fall back to the value column or whichever column or index they prefer.

How to test the changes?

(Select all options that apply)

• I've included appropriate automated tests.
• This is a refactoring of components with existing test coverage.
• Instructions for manual testing are as follows:
  1. [add testing steps and prerequisites here if you didn't write automated tests covering all your changes]

License

• I agree to license these and all my past contributions to the core galaxy codebase under the MIT license.

[mvdbeek]

The commit says Attempt to not require name column but what you're removing here seems to be the fallback definition for name. Can you start with a test case (automated or not) that breaks what you're fixing ?

[guerler]

Absolutely, thanks for looking into it. I am adding a test case right now and will update PR comment with more details.

[mvdbeek]

As a result, when searching for a column such as path, the returned index is shifted by one,

Shifted by one relative to what ? I see that https://test.galaxyproject.org/api/tool_data/twobit is missing one column relative to the fields the API claims to return, isn't that the correct bug fix ? Here you're altering all data table definitions, I don't think that's wise ?

twobit data tables do not include a name field, and since the path field is the last column, any attempt to access it for downloading data from IGV results in a failure.

I would suggest that the IGV viz throws an error there ?

[guerler]

Relative to: https://github.com/galaxyproject/galaxy/blob/dev/lib/galaxy/config/sample/tool_data_table_conf.xml.sample#L75C9-L75C39.

This does not affect all tables, only those missing the name field, specifically sift_db and twobit.

The issue is in the API layer, which inserts a pseudo column that does not exist. Alternatively, the API should throw an error and require a name column in addition to a value column, which I don't think is necessary.

[mvdbeek]

parse_column_spec is not restricted to the API layer though so you might be breaking tools. i'll have a look