rayokota opened a new pull request, #4047: URL: https://github.com/apache/avro/pull/4047
<!--
*Thank you very much for contributing to Apache Avro - we are happy that you
want to help us improve Avro. To help the community review your contribution in
the best possible way, please go through the checklist below, which will get
the contribution into a shape in which it can be best reviewed.*
*Please understand that we do not do this to make contributions to Avro a
hassle. In order to uphold a high standard of quality for code contributions,
while at the same time managing a large number of contributions, we need
contributors to prepare the contributions well, and give reviewers enough
contextual information for the review. Please also understand that
contributions that do not follow this guide will take longer to review and thus
typically be picked up with lower priority by the community.*
## Contribution Checklist
- Make sure that the pull request corresponds to a [JIRA
issue](https://issues.apache.org/jira/projects/AVRO/issues). Exceptions are
made for typos in JavaDoc or documentation files, which need no JIRA issue.
- Name the pull request in the form "AVRO-XXXX: [component] Title of the
pull request", where *AVRO-XXXX* should be replaced by the actual issue number.
The *component* is optional, but can help identify the correct reviewers
faster: either the language ("java", "python") or subsystem such as "build" or
"doc" are good candidates.
- Fill out the template below to describe the changes contributed by the
pull request. That will give reviewers the context they need to do the review.
- Make sure that the change passes the automated tests. You can [build the
entire project](https://github.com/apache/avro/blob/main/BUILD.md) or just the
[language-specific
SDK](https://avro.apache.org/project/how-to-contribute/#unit-tests).
- Each pull request should address only one issue, not mix up code from
multiple issues.
- Each commit in the pull request has a meaningful commit message
(including the JIRA id)
- Every commit message references Jira issues in their subject lines. In
addition, commits follow the guidelines from [How to write a good git commit
message](https://chris.beams.io/posts/git-commit/)
1. Subject is separated from body by a blank line
1. Subject is limited to 50 characters (not including Jira issue
reference)
1. Subject does not end with a period
1. Subject uses the imperative mood ("add", not "adding")
1. Body wraps at 72 characters
1. Body explains "what" and "why", not "how"
-->
## What is the purpose of the change
This PR fixes AVRO-4361: the collection size guard (AVRO-4241, AVRO-4300,
AVRO-4324) rejects valid data when the reader's array element or map value
schema has a larger minimum encoded size than the writer's. Examples are a
field added with a default, `int` promoted to `double`, or `null` read as
a
nullable union. The guard computed the minimum from the reader's schema,
but
the bytes were written with the writer's schema.
- **Fast reader** (`FastReaderBuilder.createArrayReader`): the writer's
array
schema is available as `action.writer`, so the guard now uses the
writer's
element type. The guard stays exact.
- **Classic and reflect readers** (`GenericDatumReader`,
`ReflectDatumReader`):
only the reader's element schema is available while reading. A new
`minBytesToRead(Schema)` keeps the existing exact bound
(`minBytesPerElement`) when the writer's and reader's schemas are equal.
That
check is computed once per reader and reset by `setSchema` /
`setExpected`.
When the schemas differ, it uses a new `minResolvedBytesPerElement`: the
minimum size of any writer's schema that resolves to the reader's.
- A field with a default counts 0.
- `float` and `double` count 1, since they may be promoted from `int` or
`long`.
- A union counts the smaller of 1 and its smallest branch.
- `fixed` keeps its size.
This bound is never larger than the exact one, so a type that was capped
as
zero-byte is still capped. And it is 0 whenever a writer's value could
be 0
bytes, so valid data is never rejected. The zero-byte allocation cap is
now
chosen from the same bound, for arrays and maps.
- When the schemas differ, the classic and reflect guards are looser than
the
exact bound, but still bound a declared count by the bytes remaining.
- `ensureAvailableCollectionBytes(Decoder, long, Schema)` is unchanged;
its
Javadoc now says to pass the writer's element schema. A protected
`ensureAvailableCollectionBytes(Decoder, long, int)` overload is added
for
`ReflectDatumReader`.
## Verifying this change
New tests, each run on both the fast and classic readers where applicable:
- `TestGenericDatumReader`
- `testMinResolvedBytesPerElement`: the resolved bound for promotions,
unions, `fixed`, and records with defaulted fields.
- `arrayOfRecordsWithAddedDefaultFieldDecodes`: the reported case.
- `arrayWithPromotedElementTypeDecodes`: `int`, `long` and `float` read
as
`double`.
- `mapWithPromotedValueTypeDecodes`
- `arrayOfNullsReadAsNullableUnionDecodes`
- `resolvedArrayRejectsHugeCount`: the guard still rejects a huge
declared
count when resolving.
- `equalSchemasKeepExactBound`: equal but distinct schema instances keep
the
exact bound.
- `TestReflectDatumReader.read_arrayWithPromotedElementType`
Without the fix, the five decode tests fail with the reported
`EOFException: Collection claims ... bytes are available`.
## Documentation
- Does this pull request introduce a new feature? no
- If yes, how is the feature documented? not applicable
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
