[
https://issues.apache.org/jira/browse/AVRO-680?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14317431#comment-14317431
]
Ryan Blue commented on AVRO-680:
--------------------------------
Overall, this patch is looking really good. I flagged a few things:
* In {{getNameForNonStringMapRecord}}, an {{UnsupportedEncodingException}} is
wrapped in {{RuntimeException}} with no reported error. I think it should be
{{AvroRuntimeException}} with a sensible error message explaining what action
failed.
* For name generation, what about using easier to read names? As long as you
don't expect to find an actual class named
{{org.apache.avro.reflect.Pair776ea00e586e8427}}, then the name can be anything
fairly unique. I'd prefer a simpler namespace since it won't actually find a
class, like "pairs", and it would be great to generate the name from the key
and value types. At least for primitives, this would be a lot more readable:
"IntBooleanPair", "LongStringPair", etc.
* This doesn't seem to produce the array-of-pairs schema when I call getSchema.
All map schemas are producing this:
{"type":"record","name":"HashMap","namespace":"java.util","fields":[]}. It does
work when I call it on Company.class, so I think it might be a bug.
* Is it possible to use the normal writeArray logic? It looks like it would be
easier to change {{write(schema,datum,encoder)}} so that a non-string map
replaces datum with its entry set, then that set is written as a collection and
each {{Map.Entry}} is passed to {{write(schema,datum,encoder}} individually.
That would eliminate the odd control flow in write and match how such maps are
handled elsewhere with the addition of {{getArrayAsCollection}}.
> Allow for non-string keys
> -------------------------
>
> Key: AVRO-680
> URL: https://issues.apache.org/jira/browse/AVRO-680
> Project: Avro
> Issue Type: Improvement
> Affects Versions: 1.7.6, 1.7.7
> Reporter: Jeremy Hanna
> Attachments: AVRO-680.patch, AVRO-680.patch, PERF_8000_cycles.zip,
> isMap_Call_Hierarchy.png, non_string_map_keys.zip, non_string_map_keys2.zip,
> non_string_map_keys3.zip, non_string_map_keys4.patch,
> non_string_map_keys5.patch, non_string_map_keys6.patch,
> non_string_map_keys7.patch, non_string_map_perf.txt,
> non_string_map_perf2.txt, original_perf.txt
>
>
> Based on an email thread back in April, Doug Cutting proposed a possible
> solution for having non-string keys:
> Stu Hood wrote:
> > I can understand the reasoning behind AVRO-9, but now I need to look for an
> > alternative to a 'map' that will allow me to store an association of bytes
> > keys to values.
> A map of Foo has the same binary format as an array of records, each
> with a string field and a Foo field. So an application can use an array
> schema similar to this to represent map-like structures with, e.g.,
> non-string keys.
> Perhaps we could establish standard properties that indicate that a
> given array of records should be represented in a map-like way if
> possible? E.g.,:
> {"type": "array", "isMap": true, "items": {"type":"record", ...}}
> Doug
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
