Closed tsufz closed 1 year ago
Sorry, fixes #339.
hi,
1) great initiative to start with the refactoring!
2) I suggest to version at the fourth digit (3.11.1.2
) for iterations in this repo, and to increase third digit only when pushing to Bioc... at least that's what I did. Other suggestions?
3) XCMS vignette fails now, it seems to have worked before. Any idea why?
hi,
1. great initiative to start with the refactoring! 2. I suggest to version at the fourth digit (`3.11.1.2`) for iterations in this repo, and to increase third digit only when pushing to Bioc... at least that's what I did. Other suggestions?
hi,
1. great initiative to start with the refactoring! 2. I suggest to version at the fourth digit (`3.11.1.2`) for iterations in this repo, and to increase third digit only when pushing to Bioc... at least that's what I did. Other suggestions? 3. XCMS vignette fails now, it seems to have worked before. Any idea why?
3.11.1.2
and push again.I need to check the code anyway again, somehow NAs are imputed in the records.
COMMENT: INTERNAL_ID 4113
CH$NAME:
CH$COMPOUND_CLASS: NA
CH$FORMULA: NA
CH$EXACT_MASS: NA
CH$SMILES: NA
CH$IUPAC: NA
@meowcat, I guess that the downstream functions expects a dataframe
? But readr::read_csv
imports a tibble()
, hence, we need to transform the imported table?
Well, for whatever reason loadinfolist
imputes now NA
to the imported table. This is the cause for the downstream problems. Will figure this out tomorrow...
@meowcat I guess, now is all okay, so read for review.
Okay, normal workflow fails, checked locally, something with the infolists due to refacorisation. However, requires update of RMassBankData as soon it works locally. In meanwhile, we could use the github package for testing.
I think the most important thing we should do before moving this to Bioc is a regression test for record creation. Basically we store the results of processing the demo dataset, strip the DATE
and MS$DATA_PROCESSING: WHOLE
entries, and check whether the generated records are still identical.
Okay, normal workflow fails, checked locally, something with the infolists due to refacorisation. However, requires update of RMassBankData as soon it works locally. In meanwhile, we could use the github package for testing.
@meowcat A sorry, the problem is that the BioC RMassBankData is not up-to-date (see latest update . I needed to change the data in the csvs to reflect changes in naming in RMassBank.
I suggested to change the source of RMassBank data package to the github main until my PR is running through and we can safely update the BioC package?
Hm but in this case should we not provide the option to update old infolists? We are breaking backwards compatibility here. I would suggest that we keep "soft compatibility" by providing updateInfolist
that rewrites the column name for old infolists.
@meowcat, I agree with the breaking argument. But, I am looking forward. I suggest rewriting the infolist in the case an old infolist is used to the new format. The dots in the names are not good, but a result of an old school read.csv()
or a wrong use of it (as @sneumann mentioned). Now is the time to change it.
According to my suggestion, the internal infofile handling is forward compatible now. Old lists are transformed to the new field format (no .
, but _
. And the spooky first column is removed (as an artefact of an incorrect use of write.csv
.)
This makes gatherCCTS more flexible and uses the full capacity of data retrieved.
I'm rerunning this job to see if that was a flaky fail
Sorry for the delay. I'm trying to go through this. Right now there is an issue with getDTXCID because it needs the API key... Do we simply omit testing this?
Sorry for the delay. I'm trying to go through this. Right now there is an issue with getDTXCID because it needs the API key... Do we simply omit testing this?
@meowcat, yes, all the CCTE requests require the API key. However, very useful to get DTXCID/DTXSID and the recent CASNR as they maintain it.
@meowcat, how to edit the code to remove the checks? Just deleting the run chunk?
@tsufz, sent a PR to your dev branch that \dontrun{}
s the CCTE code
OK, went through it. Nice work I added one more fix to the workflow (in the PR to your dev branch) to skip the CCTE lookup if no API key is present. Then let's look if the CI goes through...
A comment, I saw that we now use both httr and httr2, that seems unnecessary. Doesn't need solving today but should eventually be evened out
A comment, I saw that we now use both httr and httr2, that seems unnecessary. Doesn't need solving today but should eventually be evened out
I agree, but needs some reformatting. I will add an issue and make a note in the wiki for the coding practise.
@meowcat, I dunno what is still the issue. Your fix was properly merged in my fork...
All is fine now, right? Then: Merge!
Export and import replaced by readr versions.