A small change in S4Vectors breaks Structstrings

[hpages]

Hi @FelixErnst,

I just made a small change in the devel branch of S4Vectors that breaks the devel branch of Structstrings, sorry!

The change is a very simple one: I replaced C callable new_DataFrame with new_DFrame. See Bioconductor/S4Vectors@f2afb3e

The fix to Structstrings is straightforward:

1. Replace:

     PROTECT(ans = new_DataFrame("DotBracketDFrame", 
                                  vars,
                                  rownames, 
                                  nrows));
   

   with:

     PROTECT(ans = new_DFrame("DotBracketDFrame",
                              vars,
                              nrows,
                              rownames));
   

   in src/DotBracketDataFrame_class.c. Note the new order of the arguments!
2. Replace:

       S4Vectors (>= 0.27.12),
   

   with:

       S4Vectors (>= 0.47.2),
   

   in DESCRIPTION.

I'm not submitting a PR because I noticed that Structstrings on GitHub is not in sync with Structstrings at git.bioconductor.org, so I was not sure what would happen after you merge the PR and try to push to git.bioconductor.org. May you were going to run into some conflicts?

Anyways, the fix is really simple. My advice is that you resync your GitHub repo with Structstrings at git.bioconductor.org before applying the fix. Happy to submit a PR after you've resynced. Just let me know.

Thanks,
H.

[FelixErnst]

Hi @hpages

thanks for the heads up. I noticed it today on the bbs. However, even with the changes I get some errors locally, which are quite hard for me to pin point.

Minimal reproducable example

library(Structstrings)
dbs <- DotBracketString("((((....))))")
getBasePairing(dbs)

The rlang backtrace looks like this

<error/rlang_error>
Error in `if (!is.null(value) && (nrx != lv)) ...`:
! missing value where TRUE/FALSE needed
---
Backtrace:
     ▆
  1. ├─Structstrings::getBasePairing(dbs)
  2. └─Structstrings::getBasePairing(dbs) at Structstrings/R/AllGenerics.R:7:18
  3.   └─Structstrings (local) .local(x)
  4.     └─methods::as(x, "DotBracketDataFrame") at Structstrings/R/Structstrings-conversion.R:123:13
  5.       └─Structstrings (local) asMethod(object)
  6.         └─Structstrings:::.get_pairing(DotBracketStringSet(from)) at Structstrings/R/Structstrings-conversion.R:103:9
  7.           └─Structstrings:::.convert_char_type_to_character(ans@unlistData) at Structstrings/R/Structstrings-conversion.R:95:3
  8.             ├─methods (local) `$<-`(`*tmp*`, "character", value = `<chr>`) at Structstrings/R/Structstrings-conversion.R:80:3
  9.             └─S4Vectors (local) `$<-`(`*tmp*`, "character", value = `<chr>`)
 10.               ├─methods (local) `[[<-`(`*tmp*`, name, value = `<chr>`)
 11.               └─S4Vectors (local) `[[<-`(`*tmp*`, name, value = `<chr>`)

If I unquite the offending bit in .convert_char_type_to_character, I get an improper partitioning error through the validator.

My guess would be that the NROW function is not working in my case. Any ideas what has changed in IRanges, Biostrings or other packages, that I missed? Thanks for any pointers.

Thanks,
Felix

[hpages]

Make sure to switch the order of the nrows/rownames arguments. See 1. in my original comment.

[FelixErnst]

🤦

Thanks @hpages. Fixed it. Is on upstream/devel as well.

[hpages]

Great! Thanks