[
https://issues.apache.org/jira/browse/UIMA-2953?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13697579#comment-13697579
]
Richard Eckart de Castilho commented on UIMA-2953:
--------------------------------------------------
Having worked so long with uimaFIT, I may be a bit confused to what the needs
of plain UIMA are. I think you are suggesting that there are two use cases:
* Case 1: JCas wrappers in dedicated modules -- We assume that there are one or
more dedicated Maven modules for types (containing the type system description
and JCas wrappers). The jcasgen-maven-plugin should only generate the JCas
wrappers for the type systems within a module, not for anything imported from a
different module. The types may be defined in more than one type system file.
Hence the ant-like patterns are more comfortable than maintaining a "one rules
them all" type system file.
* Case 2: JCas wrappers in analysis engine modules -- We assume that each
analysis engine module comes with its own JCas wrappers. There is a type system
descriptor in the analysis engine module which imports type systems from other
locations. The jcasgen must follow the imports in order to find any types at
all for which JCas wrappers can be generated.
UIMA-1176 appears to be supporting case 1.
I think in both use-cases one can work with multiple top-level descriptors. In
case 2, the analysis engine description could contain multiple imports.
In any case, the decision which types an AE actually imports is a different one
from which types JCas wrappers are generated for. With respect to the latter, I
don't see how ant-like patterns replicate something that's in the UIMA XML
specs already.
> jcasgen-maven-plugin needs to support patterns
> ----------------------------------------------
>
> Key: UIMA-2953
> URL: https://issues.apache.org/jira/browse/UIMA-2953
> Project: UIMA
> Issue Type: Improvement
> Components: jcasgen-maven-plugin
> Reporter: Richard Eckart de Castilho
> Assignee: Richard Eckart de Castilho
> Fix For: 2.4.1SDK
>
>
> With the old JCasGenPomFriendly class in uimaFIT, it was possible to select
> the descriptors for which JCas wrappers should be generated using a wildcard
> pattern:
> {noformat}
> <configuration>
> <mainClass>org.apache.uima.fit.util.JCasGenPomFriendly</mainClass>
> <arguments>
>
> <argument>file:${project.basedir}/src/test/resources/org/apache/uima/fit/type/**/*.xml</argument>
> <argument>${project.build.directory}/generated-sources/jcasgen</argument>
> </arguments>
> <classpathScope>test</classpathScope>
> </configuration>
> {noformat}
> With the current jcasgen-maven-plugin, only a single descriptor file is
> possible. This is quite inconvenient for users. At least there should be the
> same possibility of specifying a wildcard pattern as in the
> JCasGenPomFriendly. Even better would be, to combine that with the commonly
> used include/exclude pattern used in many Maven plugins, e.g.
> {noformat}
> <plugin>
> <groupId>org.apache.uima</groupId>
> <artifactId>jcasgen-maven-plugin</artifactId>
> <executions>
> <execution>
> <goals>
> <goal>generate</goal>
> </goals>
> <configuration>
> <typeSystemIncludes>
>
> <typeSystemInclude>src/main/resources/types/**/*.xml</typeSystemInclude>
> </typeSystemIncludes>
> <typeSystemExcludes>
>
> <typeSystemExclude>src/main/resources/types/**/nojcas/*.xml</typeSystemExclude>
> </typeSystemExcludes>
> </configuration>
> </execution>
> </executions>
> </plugin>
> {noformat}
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
