[C++] Support for filename-based partitioning

[asfimport]

Directory-based partitioning is a feature of Arrow, but could we support filename-based partitioning?

e.g. I have a series of CSV files here all called something like foo_month_year.csv and it'd be nice to be able to read them in and then the month/year bits of the filenames then appear as fields I can filter on etc.

 

Reporter: Nicola Crane / @thisisnic
Assignee: Sanjiban Sengupta / @sanjibansg

Related issues:

• [Python] Partitioning.dictionaries in case of a subset of fields are dictionary encoded (relates to)
• [Python] Filename-based partitioning scheme (relates to)
• [R] open_dataset - add file_name as column (is related to)
• [R][Documentation] Document filename-based partitioning and filename-as-variable functionality (is related to)
• [C++] Implement ability to retrieve fragment filename (is related to)

PRs and other links:

• GitHub Pull Request #12530

_{Note: This issue was originally created as ARROW-14612. Please see the migration documentation for further details.}

[asfimport]

Antoine Pitrou / @pitrou:
cc @lidavidm @westonpace

[asfimport]

David Li / @lidavidm:
The current APIs trim the filenames before they're handed to partitioning, but assuming we can change that, we add or update the partitioning schemes to allow for this as well without too much trouble, I think. (If the filenames weren't trimmed, then it could already be done - at least in C++ - via a FunctionPartitioning.)

[asfimport]

Will Jones / @wjones127:
Supporting adding filename as a column may be a decent substitute for this.

[asfimport]

Antoine Pitrou / @pitrou:
Issue resolved by pull request 12530
#12530