[C++] JSON: improve error message when column changed type

[asfimport]

When a column accidentally changes type in a JSON file (which is not supported), it would be nice to get the column name that gives this problem in the error message.

—

I am trying to parse a simple json file. While doing so, am getting the error  JSON parse error: A column changed from string to number

from pyarrow import json
r = json.read_json('dummy.jl')

 

Reporter: harikrishnan
Assignee: Ben Kietzman / @bkietz

Original Issue Attachments:

• dummy.jl

PRs and other links:

• GitHub Pull Request #5571

_{Note: This issue was originally created as ARROW-6760. Please see the migration documentation for further details.}

[asfimport]

Antoine Pitrou / @pitrou:
Hmm, we should probably give better error messages. @bkietz

In this case, though, it seems the "length" field is first a string, then an integer. Arrow only accepts homogenous JSON, i.e. all objects in the same file must have the same schema.

[asfimport]

harikrishnan:
Ah I see. Thanks for the quick reply @pitrou . Yes definitely listing the column name here with the error message will be a saver when it comes to debugging. 

[asfimport]

Joris Van den Bossche / @jorisvandenbossche:
Indeed, a better error message would be nice. Renamed the issue to reflect this.

[asfimport]

Ben Kietzman / @bkietz:
Issue resolved by pull request 5571
#5571