I don't know any good API which does this kind of job.

I think the first approach with just one wildcard is sufficient. This makes it probably easier for you.

Replaced simple java class#getAnnotation with AnnotationUtils to support annotation inheritance.

Filtering EngineDescription if TypeCapability annotation is not present to avoid NullPointerException.

Fixes #32

If the TypeCapability is null, the IsolatedToolsTestMojo causes a NullpointerException. e.g. the IxaPosTagger does not provide any TypeCapability information and if you add the dependency you cant run any analysis.

schrieveslaach (3d6a700b) at 09 Nov 07:33

... and 1 more commit

According to my comment below/above this configuration example changes:

<configuration>
 <excludedTrainers>
 <excludedTrainer>de.tudarmstadt.ukp.dkpro.core.opennlp.OpenNlpSentenceTrainer</excludedTrainers>
 </excludedTrainers>
</configuration>

Additionally, an asterisk would be nice too:

<configuration>
 <excludedTrainers>
 <excludedTrainer>de.tudarmstadt.ukp.dkpro.core.opennlp.*</excludedTrainers>
 </excludedTrainers>
</configuration>

Or two asterisks:

<configuration>
 <excludedTrainers>
 <excludedTrainer>de.tudarmstadt.ukp.dkpro.core.**</excludedTrainers>
 </excludedTrainers>
</configuration>

I like <excludes>.

However, I noticed that the new feature you're implementing is not exactly what you would expect as a user of NLPf. I would expect that the <excludes> take care that you will not generate any NLP model. However, your filtering excludes only the testing. It also filters the engine factories which lead to unexpected behavior because as a user you create a factory and you will expect that this description will be test but it will be excluded by the filter.

I would suggest that you move the filtering to the TrainMojo because then NLPf will exclude the classes in the training process and not in the testing process. Additionally, you won't have a conflict between <excludes> and engine factories.

During writing I had the idea to name the configuration tag <excludedTrainers>.

schrieveslaach (03f4b3a6) at 23 Oct 18:47

Could you add an acceptance test too?

Okay, may be this is fine for now. But add a TODO comment.

✂

Have you tested this feature thoroughly? In the description I see <exclusions> and <exclude>. How does Maven know to resolve these configuration values? I expected <excludes> and <exclude>.

Remove.

Typo: For example, when you include OpenNLP and StanfordNLP, every available tool combination will be trained and evaluated in order to determine the best-performing NLP pipeline. However, if you do not want to train and evaluate the OpenNlpSegmenter, you can exclude it through the plugin configuration.

typo: ... all tools that are available through the Maven dependencies, therefore it is possible to exclude specific NLP tools.

BTW, thanks for setting up Gitlab. This makes it so much easier to submit patches. 👍

This MR add support to highlight the minted environment and the mintinline command, cf. minted package

schrieveslaach (866bd755) at 23 Jun 08:03

This MR add support to highlight the minted environment and the mintinline command, c. f. minted package

This MR add support to highlight the minted environment and the mintinline command, c. f. minted package