[ 
https://issues.apache.org/jira/browse/BEAM-1045?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Pei He updated BEAM-1045:
-------------------------
    Description: 
One known issue is "*" is not allowed in Windows OS.
For example, Paths.get("tempDir/*") might throw when code runs in Windows OS.
http://stackoverflow.com/questions/27522581/asterisks-in-java-path

This affecting IOChannelFactory.resolve(), toPath(), and match().

For match(), since it only requires support globs in the final component of a 
path. (local) FileIOChannelFactory could do things similar as 
GcsIOChannelFactory:
First, list all files under the directory path (this won't contain glob, such 
as *).
Then, check each returned files whether it matches glob.
In this way, glob (*) stays within Apache Beam's code.

>From match()'s javadoc:
"""
Glob handling is dependent on the implementation.  Implementations should
   * all support globs in the final component of a path (eg /foo/bar/*.txt),
   * however they are not required to support globs in the directory paths.
"""


  was:
One known issue is "*" is not allowed in Windows OS.
For example, Paths.get("tempDir/*") might throw when code runs in Windows OS.
http://stackoverflow.com/questions/27522581/asterisks-in-java-path

This affecting IOChannelFactory.match(String glob).
However, since match only requires support globs in the final component of a 
path. (local) FileIOChannelFactory could do things similar as 
GcsIOChannelFactory:
First, list all files under the directory path (this won't contain glob, such 
as *).
Then, check each returned files whether it matches glob.
In this way, glob (*) stays within Apache Beam's code.

>From match()'s javadoc:
"""
Glob handling is dependent on the implementation.  Implementations should
   * all support globs in the final component of a path (eg /foo/bar/*.txt),
   * however they are not required to support globs in the directory paths.
"""



> Windows OS compatibilities
> --------------------------
>
>                 Key: BEAM-1045
>                 URL: https://issues.apache.org/jira/browse/BEAM-1045
>             Project: Beam
>          Issue Type: Sub-task
>          Components: sdk-java-core, sdk-java-gcp
>            Reporter: Pei He
>
> One known issue is "*" is not allowed in Windows OS.
> For example, Paths.get("tempDir/*") might throw when code runs in Windows OS.
> http://stackoverflow.com/questions/27522581/asterisks-in-java-path
> This affecting IOChannelFactory.resolve(), toPath(), and match().
> For match(), since it only requires support globs in the final component of a 
> path. (local) FileIOChannelFactory could do things similar as 
> GcsIOChannelFactory:
> First, list all files under the directory path (this won't contain glob, such 
> as *).
> Then, check each returned files whether it matches glob.
> In this way, glob (*) stays within Apache Beam's code.
> From match()'s javadoc:
> """
> Glob handling is dependent on the implementation.  Implementations should
>    * all support globs in the final component of a path (eg /foo/bar/*.txt),
>    * however they are not required to support globs in the directory paths.
> """



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to