Repository navigation
Add Reducer capabilities by introducing a Reproducer that can be retrieved from TestOracle - #624
Conversation
|
Looks good, thanks! |
|
Currently, the tests are failing due to the newly-added HSQLDB implementation. |
|
Extended this with Reproducer support in SQLite3's NoREC Oracle and the StatementReducer to use it. PTAL and let me know if this looks good so far? I'm trying to figure out now how to connect the two. The problem I'm facing is that Main has what is needed to make a new GlobalState for the Reducer to use, but ProviderAdapter has the Reproducer. Here are my ideas so far:
Let me know what you think. I'm less familiar with this part of the design, so I'm not sure what the best choice is. |
mrigger
left a comment
There was a problem hiding this comment.
I'm not very sure. I think it would be reasonable to implement a copy method in GlobalState. Perhaps some care would be required to ensure that we deep-copy StateToReproduce. The other options also sound reasonable. Perhaps you can go ahead with the method you feel is most appropriate?
Btw, I'm boarding a long-distance flight and there might be some delays with my responses in the next few days. Sorry for that!
| } | ||
|
|
||
| public void setStatements(List<Query<?>> statements) { | ||
| this.statements = statements; |
There was a problem hiding this comment.
Perhaps it might be safer to defensively copy the list?
Updated all usages of TestOracle to add this type info.
… it to Reduce queries. Currently experimental and only outputs to stdout.
a177ba5 to
98529ca
Compare
|
I've updated ProviderAdapter to catch AssertionError so it can pull down the Reproducer and return it to Main. Main then uses similar logic to the previous PR to do the reducing. I've also rebased to master. It's hidden behind a flag and doesn't change any of the actual output logs, just outputs to stdout. Is this now at a good state to merge? I'll squash before merging. |
mrigger
left a comment
There was a problem hiding this comment.
Thanks, LGTM! I will use the GitHub user interface to update the branch and squash the commits.
|
Sorry, it seems that due to the new HSQLDB component, the build currently fails. |
Head branch was pushed to by a user without write access
|
I've fixed HSQLDB and it should pass all tests now. Please go ahead with the squash and merge when possible. Thanks! |
|
Test seem to be failing as apt-get cannot find Citus ?
|
|
Yes, merging the PR despite the failure. @halilozanakgul: could you please have a look at the Citus issue? |
|
Fantastic, thanks! |
Quick preview of what I'm working on. This adds the type information to return a Reproducer, but doesn't modify the
check()method. Is this heading in the right direction? If so, I can add the Reducer changes tomorrow.