[
https://issues.apache.org/jira/browse/PIG-4887?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Koji Noguchi updated PIG-4887:
------------------------------
Attachment: pig-4887-v01.patch
This is yet another parsing issue which is based on javacc.
In PIG-2597, we're planning to move away from javacc completely and move to
ANTLR but that hasn't happened for years...
With limited understanding of javacc, here's my patch.
I believe this is happening when PigFileParser's tokenizer matches the longest
match as comment. Starting from {{/*.jar}} to {{here */}}.
{panel}
register /Users/knoguchi/mydir{color:red}/*.jar;
A = load '$input' as (a1:int);
/* comment here */{color}
store A into '$output';
{panel}
In the patch, treating register as a special case and avoids next word to be
treated as a comment.
> Parameter substitution skipped with glob on register
> ----------------------------------------------------
>
> Key: PIG-4887
> URL: https://issues.apache.org/jira/browse/PIG-4887
> Project: Pig
> Issue Type: Bug
> Reporter: Koji Noguchi
> Attachments: pig-4887-v01.patch
>
>
> For a script with globbing on register followed by multi-line comments,
> parameter substitution doesn't work.
> {code:title=test.pig}
> register /Users/knoguchi/mydir/*.jar;
> A = load '$input' as (a1:int);
> /* comment here */
> store A into '$output';
> {code}
> Trying to run this would result in
> {panel}
> org.apache.pig.backend.executionengine.ExecException: ERROR 2118: Input path
> does not exist: file:/Users/knoguchi/git/pig/$input
> {panel}
> Running with {{-dryrun}}, we can see that parameter between the globbing and
> the multi-line style comment is ignored.
> {noformat}
> % pig -dryrun -param input=input.txt -param output=output.txt test.pig
> ...
> 2016-05-05 14:04:34,613 [main] INFO org.apache.pig.Main - Dry run completed.
> Substituted pig script is at test.pig.substituted
> ...
> % cat test.pig.substituted
> register /Users/knoguchi/mydir/*.jar;
> A = load '$input' as (a1:int);
> /* comment here */
> store A into 'output.txt';
> %
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
