[
https://issues.apache.org/jira/browse/BEAM-1281?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15829235#comment-15829235
]
ASF GitHub Bot commented on BEAM-1281:
--------------------------------------
GitHub user kennknowles opened a pull request:
https://github.com/apache/beam/pull/1793
[BEAM-1281] Encode GlobalWindow in one byte when it is the whole stream
Be sure to do all of the following to help us incorporate your contribution
quickly and easily:
- [x] Make sure the PR title is formatted like:
`[BEAM-<Jira issue #>] Description of pull request`
- [x] Make sure tests pass via `mvn clean verify`. (Even better, enable
Travis-CI on your fork and ensure the whole test matrix passes).
- [x] Replace `<Jira issue #>` in the title with the actual Jira issue
number, if there is one.
- [x] If this contribution is large, please file an Apache
[Individual Contributor License
Agreement](https://www.apache.org/licenses/icla.txt).
---
See JIRA and inline comment for explanation. TL;DR empty encodings are
pretty fragile. Many coders are sensitive to one byte addition, but this will
not affect anything in the main element collections, as they are within a
`WindowedValueCoder`. The motivating case is actually quite mundane.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/kennknowles/beam GlobalWindow-encoding
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/beam/pull/1793.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #1793
----
commit 219d0cc8f5094240d2eafab6d2c1579aa2c70219
Author: Kenneth Knowles <[email protected]>
Date: 2017-01-19T03:20:36Z
Encoding GlobalWindow in one byte when it is the whole stream
----
> GlobalWindow needs non-empty encoding in StateNamespaces
> --------------------------------------------------------
>
> Key: BEAM-1281
> URL: https://issues.apache.org/jira/browse/BEAM-1281
> Project: Beam
> Issue Type: Bug
> Components: sdk-java-core
> Reporter: Kenneth Knowles
> Assignee: Kenneth Knowles
> Priority: Minor
>
> Because the GlobalWindow is encoded to zero bytes, a StateNamespace built
> from the window has a stringKey "//" while the global namespace's stringKey
> is "/". As paths, these are identical, though we don't currently treat them
> as paths, quite. It isn't clear whether this is desirable. Maybe it is
> harmless but it complicates parsing and interpretation.
> For a system that actually builds hierarchical paths out of, say, some
> prefix, the StateNamespace, and a subsequent ID, the canonicalized path is
> the same so it is not possible to deserialize to the original.
> There are other gotchas associated with zero-length encodings, such as APIs
> that return zero bytes when there is no more data ready, versus returning
> zero bytes because the data is representable in zero bytes, given the context
> of knowing what type of data is expected.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
