[AVRO-642] Prettify the json dumped during exceptions while validating a writer/reader schema

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.3.3
• Fix Version/s: 1.5.0
• Component/s: java, python

Environment

Avro-Trunk

• Hadoop Flags: Reviewed
• Release Note: Print out a pretty output when dumping schema text during an exception.
• Tags: schema exceptions
• Assignee: Harsh J
• Reporter: Harsh J
• Created: 28/Aug/10 09:11
• Updated: 12/Mar/11 03:32
• Resolved: 30/Oct/10 01:31

Description

Right now the exception that's thrown when a reader and written schema don't match (while reading a file), looks like this:

Found {"type":"record","name":"Test","fields":[{"name":"stringField","type":"string"},{"name":"longField","type":"long"}]}, expecting {"type":"record","name":"Test1","fields":[{"name":"stringField","type":"string"},{"name":"longField","type":"long"}]}

While this is perfectly okay for small schema with a few fields, it gets hard to debug for complex and large ones.

The attached patch changes the exception message to look like:

Found {
  "type" : "record",
  "name" : "Test",
  "fields" : [ {
    "name" : "stringField",
    "type" : "string"
  }, {
    "name" : "longField",
    "type" : "long"
  } ]
}, expecting {
  "type" : "record",
  "name" : "Test1",
  "fields" : [ {
    "name" : "stringField",
    "type" : "string"
  }, {
    "name" : "longField",
    "type" : "long"
  } ]
}

P.s. Ideally a diff-like message would help too, but its mileage may vary. I guess with the pretty output printed, a user may easily do a diff (since they now have newlines), so us providing this is not really needed.