Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • 08:22
    macjohnny commented #10957
  • 07:59
    wing328 synchronize #10984
  • 07:47

    wing328 on master

    [php-slim4] Bump required PHP v… (compare)

  • 07:47
    wing328 closed #11039
  • 06:35
    vanjur commented #10957
  • 06:35
    vanjur commented #10957
  • 06:33
    vanjur commented #10957
  • 06:32
    vanjur synchronize #10920
  • 06:28
    vanjur commented #10957
  • 06:28
    vanjur commented #10957
  • 01:31
    ybelenko labeled #11039
  • 01:31
    ybelenko labeled #11039
  • 01:31
    ybelenko opened #11039
  • 01:02
    ybelenko closed #9200
  • 01:02
    ybelenko commented #9200
  • 01:01
    ybelenko labeled #11038
  • 01:01
    ybelenko labeled #11038
  • 01:01
    ybelenko labeled #11038
  • 01:01
    ybelenko labeled #11038
  • 01:01
    ybelenko opened #11038
Fjolnir-Dvorak
@Fjolnir-Dvorak
I did not check yet if the file ensure-up-to-date.sh does what it should do
all other languages did copy their test pom.xml files from CI/samples into samples, so that worked just fine
Is it correct that OCaml is not tested by ensure-up-to-date.sh?
Fjolnir-Dvorak
@Fjolnir-Dvorak
well, that is not so easy since I modified every generator since I patched a DefaultCodegen "bug"
If what I did was wrong it is easy to fix for me since I separated those changes in different branches / commits. I assume a rebase is possible on a pull request
Fjolnir-Dvorak
@Fjolnir-Dvorak
Would it be possible to distinguish generated from manual supplied but still needed files? At the moment the files are generated in the same folder where files are lying which were added manually. That does make a well tested change nearly impossible since there is no way to distinguish those files
Jim Schubert
@jimschubert
We could add manually modified files to each ignore file, but that doesn't solve the delete/recreate unless all generators have the extra files copied from the staging directory.
Slavek Kabrda
@bkabrda
@cbornet hey, so I added the example to https://github.com/OpenAPITools/openapi-generator/pull/3474#issuecomment-516772779 - feel free to take a look and comment
Dennis Kliban
@dkliban
@wing328 Thanks for merging the faraday PR. I actually have another commit that I put together last night for handling all the TLS related options.
@wing328
I will wait to make that PR until your new PR is merged
Dennis Kliban
@dkliban
@wing328 or should I go ahead and make my PR now?
William Cheng
@wing328
please go ahead
Dennis Kliban
@dkliban
@wing328 OpenAPITools/openapi-generator#3335 ... this is not python specific ... i ams eeing the same thing in Ruby. so even though you merged teh Faraday PR, i still can't switch to using the master branch
William Cheng
@wing328
is it due to the use of -D (system property)?
can you try switching to --additional-properties instead?
Dennis Kliban
@dkliban
@wing328 that seems to fix it
@wing328 when did this change?
William Cheng
@wing328
i think it's part of the refactoring work that we've been doing
the change (PR) should have been marked as breaking
with fallback
Dennis Kliban
@dkliban
@wing328 here is the PR that adds support for the TLS options of faraday https://github.com/OpenAPITools/openapi-generator/pull/3527/files
i had to add a new configuration template with the options that correspond to faraday
William Cheng
@wing328
OK, I'll try to merge them into one
Dennis Kliban
@dkliban
@jimschubert I am ready to talk about adding a 'toDisciminator' method to the CodegenConfig interface
oops
wrong link
Dennis Kliban
@dkliban
and i looked at the CodegenConfig interface. I see it has toVarName and toParamName methods
and i see where those are called later ... but i am not quite sure where you were thinking toDiscriminator would be called ... and would i need to implement a custom discriminator?
William Cheng
@wing328
Hi all, we've collected some feedback from the community and would like to announce we're switching the chatroom from Gitter.im to Slack. Please join our Slack channel via https://join.slack.com/t/openapi-generator/shared_invite/enQtNzAyNDMyOTU0OTE1LTY5ZDBiNDI5NzI5ZjQ1Y2E5OWVjMjZkYzY1ZGM2MWQ4YWFjMzcyNDY5MGI4NjQxNDBiMTlmZTc5NjY2ZTQ5MGM and we'll only answer questions in the Slack channels moving forward. We'll keep this Gitter chatroom for a few months and remove it when there's no activity. Thank you.
Fjolnir-Dvorak
@Fjolnir-Dvorak
My suggestion would be on that that we keep this gitter here as support chat. Slack seems very inappropriate for that (no one is using it and it requires authentication without GitHub). GitHub is widely used and accepted. I do have to agree with that that perhaps slack could be better for development?
The problem I have with moving to slack: I do not report any bugs if I have to login into any platform. If they want me to login into their system, they are not interested in getting bug reports so why should I bother helping them?
And sometimes I am reporting bugs via gitter. Writing bugreports is a science for itself sometimes. It's way to complicated for just reporting that something is wrong if I am not interested in fixing the issue myself
Jim Schubert
@jimschubert
The problem with Gitter is that notifications often don’t work. On both desktop and mobile, I frequently have to close the client in order to see new messages. This makes having a “conversation” pretty difficult. There’s also no support for threaded comments, and editing comments on mobile is not exactly functional. Gitter isn’t a main product focus of GitLab, and as such it’s been years and these bugs are still not fixed. Makes one wonder if unanswered messages are just dropped.
Colton Kopsa
@coljamkop
Is there a way to do selective generation from the CLI?
Fjolnir-Dvorak
@Fjolnir-Dvorak
There is a tool written in JavaScript, sadly. It's called openapi-extract. The generator itself is not selective in any way...
Colton Kopsa
@coljamkop
In general what is the best way to update generated clients? It looks like the maven plugin has some ability to be selective.
I'll take a look at openapi-extract in the mean time
Fjolnir-Dvorak
@Fjolnir-Dvorak
The best way to update the generated clients is to not edit them, delete them and regenerate them. Models which are no longer needed will not be deleted. Controllers which are not longer needed are not deleted. Are you only asking for one specific Java generator or for other languages, too?
Colton Kopsa
@coljamkop
the java11 client from the generator
I don't have many edits, just updating the pom on a few things to get it to actually build, and adding a Jenkinsfile and such
I just didn't know if I was doing things incorrectly by basically just doing the delete and regenerate approach
Fjolnir-Dvorak
@Fjolnir-Dvorak
It should work. The generator is generating the whole project and does not rename anything if you did not update the generator in the meantime, so everything should work
Colton Kopsa
@coljamkop
Okay, I'll give it another shot
Colton Kopsa
@coljamkop
Yep, looks like it worked. Out of curiosity though, there are stubs for unit tests in the client code and on the server side of things, there's going to be quite a bit of implementation that goes into hooking the server up to databases and stuff. When regenerating that code, do the tests and implementation code get wiped out every time? That was my experience, but I could have been doing something wrong.
Fjolnir-Dvorak
@Fjolnir-Dvorak
It would be easier to not touch any file which was generated. If you have to change those either you utilize git diff every single time or you put those in .openapiignore so that those will net be regenerated, and also not longer extended, corrected, etc.
For python I wrote myself a full wrapper and include the wrapper as a new pythonpath so it does not destroy any of my code. Since I had to touch the controller which was autowired (and got destroyed every time I made an api update) I created a monkeypatch which takes a manually created implementation class, patches all implemented functions into the generated code and repaires every single other endpoint to 501 "Not Implemented" and removed the garbage 200 "Do some magic". One of the only servers I know about which just works without editing generated code and destroying all dreams about maintainability is the Haskell generator. The clients should be a lot less problematic though...