epam
Generates OSGi bundle and source bundle.
This change adds OSGi support to parso and bumps the version number to 2.1.0-SNAPSHOT.
In another branch (aborg0/time_handling), I have merged some improvements from the KNIME's version of parso. In case you like them, I would like to ask your opinion about another change, which I could also merge to that branch: tracking the line number and improved error reporting based on those (feel free to check the KNIME parso lib, look for the ParsoException and the byteArrayWithData method).
@aborg0 I'm wondering if your improvements fix the issues with DateTime formats. See my pull request with a fix for a specific time format issue I ran into.
@aborg0 Never mind ... your fork surfaces the bufferunderflow exception as well...
@srozsnyai You are right, the time-related patches are not containing your proposed fix as I have no SAS7BDAT file which I could use to reproduce/fix, KNIME customers have not yet reported a problem about that.
Honestly I'm a little bit reluctant to make these type of changes... From my perspective parso should be a simple jar, and if an osgi bundle is required, either a separate project (e.g. parso-osgi-bundle) should be created or the parso jar should be converted to osgi bundle using one of the standard approaches.
Please let me know what you think so we try to satisfy everyone :)
An OSGi bundle is just a simple jar -you can check it if you try the usual packaging on this branch- with some extra version information (the only* difference would be that on maven central repository not "jar" would be the text on the download button, but "bundle", those who are not familiar with it, would hardly recognize). Using gradle as I used is the standard way to generate an OSGi bundle. ;) (Though I have to admit most projects do not care to make the source jar a bundle too. If that is too much, I can remove that.)
The reason why I think this should be a bundle and not a separate project: when epam signs the jar, the bundle should be also signed, while if the owner of the OSGi project have no access to the key used for signing, another should be used, so authenticity is lost.
*: There is one more thing, which might be a downside: semantic versioning should be used for releases, so in case there are breaking changes in the public interface -I doubt there will be many-, the version number have to be incremented.
@printsev would this change be more acceptable if the implementation package was not exported? I think that can be done with OSGi services (probably a nicer solution). Maybe you require unit tests for OSGi functionality too? Obviously a choice of not supporting OSGi is also acceptable, we can generate bundles from released parso builds quite easily (either with maven or gradle), so it is not a problem for us, just thought others might benefit from an official OSGi bundle too (which would still allow using parso as a regular jar too).
Have you had a chance to take a look at the other changes in the time_handling branch? Do you think that would be useful for parso, or we should move them to a separate processing step? (I think without the time handling changes the encoding and the improved error handling (which is not yet added, just a precursor of it - a new exception type should contain the row number too) might be useful.)