New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Create package hierarchy #13
Conversation
This is perfectly OK. You should provide vocabulary object, but instead you have string "en" in the configuration. I am not sure where did you get this configuration file, but it seems to me it is based on one of the proposals how the configuration should like but was never implemented. |
It must have been implemented, I ran it quite a few times. But I see where it differs from |
I've just pushed a fixed config. |
... and now I have to rebase this. I'm not quite sure what the correct workflow is: should you have posted this as a pull-request for my fork? Thanks anyway. |
Now it fails on
Also, I'm not sure we will be able to run any meaningful tests on Travis, because it has lower version of glibc than TF is compiled with. |
I've found a way to run tests on Travis. Strangely enough, it shows different error than my local computer. Ad correct workflow: this should have been a branch here, not in my fork, then you would be able to push to it. Mea culpa. |
OK, now it arrived at the same error. Can you take a look, @jlibovicky? I honestly cannot tell whether this is another error it my ini file or your hungry config-builder ate another exception. Edit: Even though it probably only eats import errors, so this should be something else. |
This is because of @jindrahelcl recent change in the dataset objects contructions. All series names must no begin with "s_". I always forget about this one. Btw. @jindrahelcl, did you have a good reason to make it this way? |
I changed it when I was adding the preprocessor option to the dataset config. It seemed simpler to me just to prefix names for dataset series with s_, than to have some restricted keywords that nobody knows or not using underscores in option names. But I agree it's not the best solution either. Perhaps we could enable the options to have dicts as parameters. Then in dataset, you would have an option |
I am aware that I changed also the usage of the preprocessor. Now, it is used on the whole dataset, which is not sufficient, we need to be able to use also preprocessors on a subset of series, jointly or independently. It should also be done in a way that user can stack multiple processors on top of each other. |
I don't see anything beginning with "s_" in any |
OK, I solved it. The next error message is very confusing again. First it says |
Also, why are the vocabularies empty? |
Solved. Now it says that it's missing series. |
OK, I get the joke, we write "s_" on the left side, but not on the right. Seems like I'm finally dealing with real errors. |
This is an interesting error, seems like a bug in postprocessing. |
Finally, the test configuration works and we can run a real test on Travis. This should prevent things like #21 from happening. I have not yet corrected all the imports (I intend to do that as I work on #6), so you may still run into import errors, but fixing them is easy enough, so I consider this ready to merge (after we move away the last commit from master, as discussed in #21). What do you think, @jindrahelcl, @jlibovicky? |
One more thing - the |
What's the problem with the I agree that the I see no other reason not to merge this (other than #21, but that can be solved later). So unless someone objects, I will merge this tonight. |
Ok, it looks good now. |
This does not work yet. Also, Travis will not run it, because glibc.
More importantly, this fails with
I do not know what to do with this. My guess is, that the horrible monster of configuration-builder ate some useful exception somewhere.