haven update 2.5.4 to 2.5.5 cannot read .sav-file anymore

[MarcusPenthin]

Dear developers,

after updating this package to the newest version, I am not able to read in my datafile (https://osf.io/download/j6vah/) anymore, in version 2.5.4 the R-Code (see below) worked perfectly. Even if the file is on my local disc. Installing going back to version 2.5.4 manually, everything works fine.
With package "foreign" there is no issue.

This is the reprex:

library(haven)
data_notworking <- read_sav("https://osf.io/download/j6vah/",
                  encoding = NULL, user_na = FALSE, col_select = NULL, skip = 0, n_max = Inf, .name_repair = "unique")
#> Error: Failed to parse C:/Users/semapent/AppData/Local/Temp/Rtmp2v3WlW/filebc185132513: Unknown error.

library(foreign)
data <- read.spss("https://osf.io/download/j6vah/",
                  use.value.labels = FALSE, max.value.labels = Inf, trim.factor.names = FALSE, to.data.frame = TRUE,
                  trim_values = TRUE, reencode = NA)

Other files are working fine. Don't understand what the problelm might be. Is there something corrupted in my datafile?

Kindly,
Marcus

[maxecharel]

Exactly the same problem here, with the same 'unknown error', as well as the same 'workaround': back to 2.5.4, it works as expected

[rolsen-dft]

Hi, I managed to fix this issue by reverting to an older version of SPSS. I have just made a post about this (sorry, I did not see this post before I made my own).

[maxecharel]

Hi, I managed to fix this issue by reverting to an older version of SPSS. I have just made a post about this (sorry, I did not see this post before I made my own).

Hi @rolsen-dft, as discussed in @gorcha 's ReadStat PR (many thx @gorcha for the PR):

From #788 - the new MR set changes only support ascii characters in the set name in the ragel parser, but SPSS uses the file code page/UTF-8 for these.

This PR updates the parser to allow for non-ASCII characters, and runs text from the MR set through readstat_convert() to make sure the character encoding comes in correctly.

So according to what you said regarding downgrading SPSS, that would imply that SPSS previously used ASCII for the MR set and is now using UTF8. Is that right?

Downgrading SPSS to regenerate the binary can only be considered as a temporary workaround, and one that could be seen as out-of-reach for some of us (depending on the provider of the serialized data or your IT division). While it is interesting to have it mentioned here, maybe could you consider closing the issue you've created since the issue by itself is redundant, in order notably to put more emphasis on this one?

[gorcha]

Hey all,

Confirming this is because any characters outside of the standard ASCII character set are not supported by the new MR set feature in readstat.

It may potentially be a different issue that you're having though @rolsen-dft - the original file that @MarcusPenthin shared just has standard accented characters, which are still supported by pre-UTF versions of SPSS, so I'm a little surprised that saving with an older version of SPSS fixed it. What are the versions that you used?

Are you able to share a file by any chance (since this issue is with metadata a version of the file with all observations deleted could still trigger the issue)?

[gorcha]

This should be fixed in the dev version of ReadStat - I've updated this in #794.

@MarcusPenthin your reprex is now working fine. @maxecharel and @rolsen-dft could you please have a look and let me know if you're still getting the same error?

[maxecharel]

This should be fixed in the dev version of ReadStat - I've updated this in #794.

@MarcusPenthin your reprex is now working fine. @maxecharel and @rolsen-dft could you please have a look and let me know if you're still getting the same error?

Hi @gorcha, thank you for your PR. Since not merged yet, I targeted the commit:

pak::pak("https://github.com/tidyverse/haven/commit/4915bb5")
devtools::session_info()$packages |> filter(package == "haven")
## package * version    date (UTC) lib source
## haven   * 2.5.5.9000 2026-01-24 [1] Github (tidyverse/haven@4915bb5)
test <- haven::read_sav(file = "~/test/savserial.sav")

Unfortunately, I get the following error:

Error: Failed to parse /home/[user]/test/savserial.sav: A multi-response set record is invalid.

Edit: to see the glass half-full, at least with your PR the error is not Failed to parse, unknown error anymore. Thank you for your effort, do not hesitate to make me dig further

[gorcha]

Hey @maxecharel, thanks for checking it out! Are you able to share the file that you're getting the error with so I can dig a bit deeper?

The MR set functionality is pretty new so it may be something else.