Scope

Seems like this turns out to be a UTF-8 byte order mark (BOM) issue. Although it was resolved back in #12970 for csv-parser, csv-parser was then changed to papaparse in #19739, hence BOM isn't being handled again.

Related issues in PapaParse's repo:

• Unicode BOM messes up first property name mholt/PapaParse#372
• First and Last CSV column issues mholt/PapaParse#855

With the sample.csv reproduction file provided in #21727 which contains BOM (if you open it with vscode or notepad, the bottom right corner should say UTF-8 with BOM), as well as using the transformHeader config option as a quick test:

transformHeader: (header) => {
    console.log(header);
    return header;
},

That allows us to identify the problem, which is the first column always_ignored actually gets processed as always_ignored (notice the visible leading space):

This causes the import process to attempt to import to a technically non-existing column, hence the missing data for users that are using CSV with BOM for their first column every time.

This PR opted to use a simplistic .trim() as brought up in PapaParse's issues, but other potential alternatives could be:

• only trimming it when index === 0 to make it clearer we're cleaning up just the first column

• Only cleaning up BOM specifically which a slightly more elaborate code such as:

  transformHeader: (header) => {
      if (header.charCodeAt(0) === 0xFEFF) {
          return header.substring(1);
      }
      return header;
  },

What's changed:

• added transformHeader papaparse config for CSV imports to trim the headers

Potential Risks / Drawbacks

• Maybe a tiny overhead for trimming the headers

Review Notes / Questions

• Whether the simple trim() is acceptable, or other alternatives mentioned above are more preferred since they are more specific in terms of code readability/maintainability

Fixes #21727