Remove local library attachment in module run script

[anthonysena]

Aims to fix #96 by removing the local library attachment and to simply use the libraries recorded in the module lock file. At the moment this assumes the modules have all of the necessary dependencies available to run the auto-generated script.

[codecov]

Codecov Report

Attention: 3 lines in your changes are missing coverage. Please review.

Comparison is base (00d0c00) 78.05% compared to head (dd3634b) 67.69%.

Files	Patch %	Lines
R/RunModule.R	0.00%	2 Missing ⚠️
R/ResultModelCreation.R	0.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##           develop     #107       +/-   ##
============================================
- Coverage    78.05%   67.69%   -10.37%     
============================================
  Files            8        8               
  Lines          793      780       -13     
============================================
- Hits           619      528       -91     
- Misses         174      252       +78     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[azimov]

I assume the redshift tests are failing?

[anthonysena]

Yes, I need to debug this but for now I've disabled the RedShift tests. It appears to be related to the way I'm downloading the jdbc driver and the tests cannot find the driver for that one for some reason.

[azimov]

Will keyring and ParallelLogger still have issues? They are called in other places in this script and that would be harder to fix? I've been trying to break this and it seems ok though

[anthonysena]

That is still possible - the reason why these work now is because they are included in the module renv.lock environment. Assuming the module is restored properly AND has these dependencies it will work. I'd like to formalize these type of dependencies as mentioned in #69.

[hmorgancooper]

Hey :) I was having issues running Strategus due to the renv not being able to find packages in my local library. I think this issue is the one to fix that so I'm posting this here but sorry if this isn't the right spot.

I tried running the develop branch of strategus to test the fix and I am now getting an error with the CohortGeneratorModule.

It cannot find the MetaData.json inside /instantiatedmodules/CohortGeneratorModule_0.2.1. I took a look in that folder and it is empty apart from .Rhistory.

Does anyone have any ideas what might be going on?

[anthonysena]

Hi @hmorgancooper - I am guessing a bit here but the fact that the folder exists but it is empty (apart from the .Rhistory) leads me to think that the code was unable to download the CohortGeneratorModule from GitHub. I'm not sure of your environment but is that a possibility?

[hmorgancooper]

It could be that!

Does the Strategus develop branch attempt to download the module from github if it's not present? Or do I need to download it and have it in my r environment? I do have CohortGenerator already installed but I'm assuming I need CohortGeneratorModule as well?

Another point, I originally had the main branch of Strategus which I removed and then replaced with develop to test the new change. Is this likely to have caused any weirdness with the other modules or am I ok to try that?

Thanks!

[anthonysena]

Hi - Strategus will attempt to download and install the module if it is not found in the INSTANTIATED_MODULES_FOLDER. The develop branch of Strategus will go one step further to verify that the renv dependencies defined in the renv.lock file that is present in the root of the module folder before it attempts to run the Strategus study.

Removing/reinstalling Strategus will not remove any prior modules that were previously installed. Hope this helps and please let me know if you are able to get the modules installed. Thanks!

[hmorgancooper]

Hey Anthony, thanks for the help :) The error message has changed which means I'm making progress!

I deleted the empty CohortGeneratorModule from the instantiated modules folder and it recreated it successfully (I think). The CohortModule package ran as far as I can tell. It threw up some warnings about not being able to rename some files because they did not exist (cohort_inclusion.csv, cohort_inc_result.csv etc) but it did finish.

I'm now erroring out in the CohortIncidenceModule.
I have a warning saying the renv is out of sync - apparently OhdsiRTools and Strategus are not available. When I run renv::restore() it's telling me the library is already synchronized with the lockfile though.

The actual error now is a java SQL error - Syntax error: Expected ")" but got keyword GROUPS at [59:5].

I am using BigQuery SQL which has caused some issues in the past. I got this 'hint' in the errorReportSql.txt
--HINT DISTRIBUTE_ON_KEY(subject_id)

For some context I am trying to run the HowOften Study (I can ask about this issue on that github page too if that makes more sense)

Thanks!

[hmorgancooper]

I managed to get the whole study to run! Thank you for the help :)

In case anyone runs into a something similar here is the link to the issue I raised on HowOften.
ohdsi-studies/HowOften#14