[
https://issues.apache.org/jira/browse/THRIFT-342?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12912493#action_12912493
]
Bryan Duxbury commented on THRIFT-342:
--------------------------------------
This just changes our implementation of "set" in PHP with an array, right? If
there's no reliable set construct, this seems like the next best thing.
Unless there are any objections, I'll commit this later today.
> PHP: can't have sets of complex types
> -------------------------------------
>
> Key: THRIFT-342
> URL: https://issues.apache.org/jira/browse/THRIFT-342
> Project: Thrift
> Issue Type: Bug
> Components: PHP - Compiler, PHP - Library
> Environment: SVN trunk r743881
> Reporter: David Sklar
> Attachments: thrift-342.patch
>
>
> A setup like this:
> struct alice {
> 1: string bob
> }
> and another like this:
> struct charlie {
> 1: set<alice> david
> }
> causes problems because the generated PHP code looks like:
> ==
> case 1:
> if ($ftype == TType::SET) {
> $this->david = array();
> $_size0 = 0;
> $_etype3 = 0;
> $xfer += $input->readSetBegin($_etype3, $_size0);
> for ($_i4 = 0; $_i4 < $_size0; ++$_i4)
> {
> $elem5 = null;
> $elem5 = new alice();
> $xfer += $elem5->read($input);
> $this->david[$elem5] = true;
> }
> $xfer += $input->readSetEnd();
> } else {
> $xfer += $input->skip($ftype);
> }
> break;
> ===
> Using objects as array keys makes PHP cranky and the values can not be
> properly set. I think the solution to this is either:
> 1. Document that the PHP bindings do not support sets of complex types.
> (boooo!)
> 2. Modify the generated code so that the values are stored in array values
> not keys. (With some additional checks to ensure uniqueness, or perhaps just
> using spl_object_hash($theObject) as the array key.
> This seems similar to the issues raised in THRIFT-231 and THRIFT-162
> $this->david[$elem5] = true;
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
