Fix labeled string values bug

[jamesalster]

PR proposing a fix to issue #52, with a brief test of the file read. I'm new to this repository and the topic, so I may have made some blunders. I haven't worked out if we would ideally read in string columns with appropriate string lengths, nor how to do that.

A file to recreate the issue is added at data/string_labeled_value_eg.sav. Reading this in with haven, and this branch, gives the correct string value => string label mapping, showing the regions of Belgium (in this case). But on main, the regions of Belgium are not read in.

Tests run fine on my machine.

[junyuan-chen]

Well, the reason why value labels over strings are not allowed here (intentionally) is because libreadstat requires the values attached with value labels to be either Int32 or Char from what I remember. It's possible that value labels may be allowed for string columns (in SAS or SPSS), but I will want to take a look at what haven does to make this work.

[codecov[bot]]

Codecov Report

:white_check_mark: All modified and coverable lines are covered by tests. :white_check_mark: Project coverage is 99.92%. Comparing base (e179392) to head (1b036df).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #53   +/-   ##
=======================================
  Coverage   99.91%   99.92%           
=======================================
  Files          11       11           
  Lines        1249     1254    +5     
=======================================
+ Hits         1248     1253    +5     
  Misses          1        1           

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

:rocket: New features to boost your workflow:

• :snowflake: Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jamesalster]

Right, I see! Thanks, I'm not familiar with libreadstat. From a user point of view, I would say that handling the string labels is definitely expected behaviour, since they do appear from time to time in mainstream datasets (the file I used there is an excerpt from the European Union's Eurobarometer 2010).

But I of course defer to you on implementation, and I'm happy to contribute more time to implement things another way, if you suggest a direction.