[test] index test updates OP#318 #6033
Conversation
…ne, ngram, and range indexes see eXist-db#787.
|
In terms of debugging tests and fixing issues in the actual underlying Java code, this is a bad idea. XQSuite is horribly convoluted when trying to understand the real problems that are in the Java code. This will make the developers work much harder. You would have been better going in the opposite direction and converting the XQSuite tests into the older XML format, it's runner implement is much simpler and much easier to debug the Java code when it executes |
unskip ~17 tests which are passing since they were added decouple tests from classpath based sample jar increase test isolation to avoid setUp tearDown from differnt specs interfering with each other
so they no longer depend on state from xml testset
|
execution times for last test run on linux develop and here are Turns out some of the existing tests were relying on state from the xml specs or from previous tests, which mavend exaggerated on CI. Sharing state does not strike me as conducive to a good developer experience. @adamretter the horrible convolutedness of a mix of xml and xq tests while working on the underlying java code is what motivated this PR. The work won't stop here. |
But switching to XQSuite makes this much much worse. Are you doing to instead switch all the tests to the XML format instead then to improve this? |
| : @param $group group name: smp-indexed | smp-dropped | sip-indexed | sip-dropped | ||
| : @return xs:string "group: count" e.g. "smp-indexed: 5" | ||
| :) | ||
| declare function unic-smp-l:lucene-finds-count($group as xs:string) as xs:string { |
There was a problem hiding this comment.
This here could be rewritten to a custom assertion
There was a problem hiding this comment.
yes and it might in a future refactoring of the lucene tests, but this is the closest to the issue report. So I d rather leave this commit as is and focus on a larger refactoring of the entire Lucene suite, including this spec.
|
@line-o so good to merge then? |
4 participants
while going though old issue and checking their test coverage I found the mix of existing xqsuite test in xml testSet format, and xquery format irritating. This now unifies on XQ format, for easier portability, and better test isolation. I could also unskip a few test which hadn't been running since the spice girls, but which are now passing. All test from xml have been refactored keeping their syntax and intention as intact as possible.
What actually go me going are the first 4 commits. These are pending tests for open issue, confirmed to still exist in our code base since long before the spice girls.
I expect further refactoring of the indexing tests and more additions, so it would be good to get this merged as the basis for that. Focusing on:
truewhere more expressive assertions, makes it easier to see what went wrong when something does go wrong.see #2393