Closed zifeishan closed 8 years ago
Could you link to the exact change in the future? You're probably talking about this change, right? That is supposed to run with HazyResearch/deepdive@bebd95665b6292fbf35b9a144ef17eecd6326ee4 which hasn't been merged into DeepDive's master yet. Things will get sorted out soon, but these are going through major refactoring, and sorry if my bugfix #87 was mixed up in the middle of these. I think we should keep a maintenance branch for ddlog as well and rebase the bugfix onto there.
In general, ddlog isn't supposed to work with arbitrary version of deepdive and vice versa. It's a submodule, hence only a specific combination of those two will work correctly. Since ddlog compiler is going through a series of refactoring alongside lots of changes in deepdive compiler, things may stay a little bumpy if you mix and match the two on your own.
The integration tests are already done fully from the deepdive side, and I think it doesn't make sense to repeat the outer tests within a nested submodule.
@zifeishan I created a v0.8.x branch in ddlog which contains just the bugfix. deepdive master/v0.8.x will be updated to use this soon. Please use this instead of master :)
Recently a change in ddlog breaks the compiled path of
udf
(by using relative paths instead of absolute paths), and we ran into errors when we run end-to-end applications (udf not found). It seems that we don't have good end-to-end coverage for ddlog applications.We probably want to have tests running example applications from ddlog compilation to deepdive execution, to make sure everything works together correctly.