EDIT

I implemented a new class eu.etaxonomy.cdm.common.URI and use this in model but also at other places like import sources.

Unfortunately, simply subclassing java.net.URI was not possible as it is final.

A method getJavaUri() exists to transform the above class to standard java.net.URI.

Currently there are still 2 open issues:

• httpinvoker: within httpinvoker there is still a ClassCastException and therefore TaxEditor fails to build. First I thought this is because CdmModelCache needs to be build anew, but running the script does not change the model cache file so this does not seem to be the reason
• RdfViewTest.testMarshalRdf does not run successfully and therefore I set it to ignore. The children type dc:identifier/ do not seem to be build correctly.

Katja, can you please check if you can find out what the httpinvoker problem is. I currently can't start the TaxEditor from the IDE for debugging.

This ticket seems to be done or needs to be splitted into an open issues ticket. It can be moved back to 5.19(?) afterwards.

Andreas Müller wrote:

In #9111#note-1 AK suggests to also encode fragment and query parts. This has not been implemented in the original code in #9111 which has been copied to cdmlib.

I think we should also implement it for fragment and query. Probably not for authority and scheme?

Also the handling of backslash needs to be discussed. In the original version backslash was encoded, too. This is probably not wanted, at least not for the path part. Currently a URISyntaxException is thrown if path contains a backslash.

Did you implement this already?

Katja Luther wrote:

Andreas Müller wrote:

In #9111#note-1 AK suggests to also encode fragment and query parts. This has not been implemented in the original code in #9111 which has been copied to cdmlib.

I think we should also implement it for fragment and query. Probably not for authority and scheme?

Also the handling of backslash needs to be discussed. In the original version backslash was encoded, too. This is probably not wanted, at least not for the path part. Currently a URISyntaxException is thrown if path contains a backslash.

Did you implement this already?

No, nees to be finished

Andreas Müller wrote:

Katja Luther wrote:

Andreas Müller wrote:

In #9111#note-1 AK suggests to also encode fragment and query parts. This has not been implemented in the original code in #9111 which has been copied to cdmlib.

I think we should also implement it for fragment and query. Probably not for authority and scheme?

Also the handling of backslash needs to be discussed. In the original version backslash was encoded, too. This is probably not wanted, at least not for the path part. Currently a URISyntaxException is thrown if path contains a backslash.

Did you implement this already?

No, nees to be finished

Should I do this? Because you worked on this ticket.

I only moved the original algorithm to the user type. The remaining issues are about the algorithm. As you implemented the algorithm I thought you will do this.
Or at least give hints why e.g. fragments and query parts were not yet encoded and why backslash was encoded. I didn't want to change the algorithm without feedback on these issues.
Maybe also Andreas K. can give feedback on this.

This still needs feedback so I move to current milestone. Once it is fixed we can move to 5.19 where most of the work was done or keep it in current milestone.

Implemented encoding of fragment part, query part was already implemented. @AK could you have a look and give feedback whether this implementation is ok?

According to the URI and URL Schemes defined in the corresponding RFCs:

URI = scheme ":" hier-part [ "?" query ] [ "#" fragment ]

       |------------------ Schema-spezifischer Teil ------------------|

 https://max:muster@www.example.com:8080/index.html?p1=A&p2=B#ressource
 \___/   \_/ \____/ \_____________/ \__/\_________/ \_______/ \_______/
   |      |    |           |         |       |          |         |
Schema⁺   | Kennwort      Host      Port    Pfad      Query    Fragment
       Benutzer

There is always only one fragment. So it makes not much sense to handle multiple fragments as in our code.
A better algorithm would be to split the path+query+fragment string at the first # character and to tread the
right part as fragment.

BTW: There is no test method in eu.etaxonomy.cdm.common.UriTest to test URI strings with fragment.

Thanks for your comment, I adapted the code and added a test.

Please add a testcase with multiple fragments, like https://max:muster@www.example.com:8080/index.html?p1=A&p2=B#ressource&#another-frag?

Applied in changeset cdmlib|09da5a1bd580baae6490435bc52dddd6d2aab353.

Added a testcase for more than one fragment, the rest was already reviewed, so we can close this ticket.

This was fixed in 5.30 already