#general
- w
witty-crayon-22786
10/26/2017, 4:40 PM@fancy-queen-20734: great. the other bit you should be able to remove then is the linear search through deps - f
fancy-queen-20734
10/26/2017, 4:40 PMI think that should resolve the things I was worried about, but I’d appreciate a counter example to fill out the test cases - f
fancy-queen-20734
10/26/2017, 4:40 PMyeah - w
witty-crayon-22786
10/26/2017, 4:40 PM - w
witty-crayon-22786
10/26/2017, 4:40 PMbecause you already have that mapping - f
fancy-queen-20734
10/26/2017, 4:42 PMhm. I don’t though. That’s covering the original exports. Another option here would be to do export injection where the dependency injection is happening - f
fancy-queen-20734
10/26/2017, 4:42 PMbut I don’t have a good story for updating payload fields - f
fancy-queen-20734
10/26/2017, 4:42 PMI felt like that would be a bad idea - w
witty-crayon-22786
10/26/2017, 4:43 PM@fancy-queen-20734: what do you mean it's covering the original exports? - w
witty-crayon-22786
10/26/2017, 4:43 PMit's covering all codegen exports, which are the only ones you need to do that check for - f
fancy-queen-20734
10/26/2017, 4:44 PMSo, the purpose of that part is to allow for removing that code from Target. - f
fancy-queen-20734
10/26/2017, 4:45 PMmaybe I could merge those searches together to do one pass of the dependencies rather than n - w
witty-crayon-22786
10/26/2017, 4:45 PMi don't see why export injection is necessary... exports are a local-only concept - w
witty-crayon-22786
10/26/2017, 4:45 PMto make this concrete: - w
witty-crayon-22786
10/26/2017, 4:45 PMthe model change i have been suggesting is that the complete list of exports and dependencies is added to a target at injection time - w
witty-crayon-22786
10/26/2017, 4:46 PMand then you're done touching that target. - f
fancy-queen-20734
10/26/2017, 4:46 PMyes, but an export on a codegen target is a little special because you’d expect it to work like a dependency on a codegen target, ie if the referenced target is also a codegen target, it should inject an edge to the synthetic target - w
witty-crayon-22786
10/26/2017, 4:46 PMyou still have to mutate its existing (ie, non-codegen) dependents, but - w
witty-crayon-22786
10/26/2017, 4:46 PM@fancy-queen-20734: no - w
witty-crayon-22786
10/26/2017, 4:47 PMexports do not cause additional edges to be injected - w
witty-crayon-22786
10/26/2017, 4:47 PMunless i'm missing something large. - w
witty-crayon-22786
10/26/2017, 4:47 PMthe edge exists because of the dependency - w
witty-crayon-22786
10/26/2017, 4:48 PMand if a codegen target depends on a codegen target... - w
witty-crayon-22786
10/26/2017, 4:48 PMwhen you create the dependent, you can give it a complete exports list. - w
witty-crayon-22786
10/26/2017, 4:48 PMbottom up. - f
fancy-queen-20734
10/26/2017, 4:49 PMhm. I think the way I dealt with the target.py bits is wrong because I’m not / we can’t inject export references to synthetic targets from non-synthetic ones. I need to revert that bit - f
fancy-queen-20734
10/26/2017, 4:50 PMSo, currently code gen targets inject edges between their new synthetic targets and the dependents of the codegen target. Those new dependencies, if there were matching exports need to include injecting the exports for the new dependencies - f
fancy-queen-20734
10/26/2017, 4:51 PMbut we don’t allow for injecting exports right now, so we have to do what Target.exports is doing - w
witty-crayon-22786
10/26/2017, 4:51 PM@fancy-queen-20734: only for codegen dependents - w
witty-crayon-22786
10/26/2017, 4:51 PM...right?