Written dta file with version=15 not readable in Stata 17 Basic Edition

[ofajardo]

If writing a dta file setting version as 15, when opening it in Stata 17, the user gets the error:

This .dta file format was created by Stata/MP and has more variables than your Stata can handle.

The file can be opened however if the version is set to 14.

Looks kind of similar to this

Original report

[gorcha]

FWIW, this actually appears to be correct behaviour. Stata itself uses a different version numbering scheme. Current Stata versions use format 118 for most files, and 119 when there are more than 32,767 variables.

These are mapped to version = 14 and 15 respectively in pyreadstat (we do the same in haven), when actually they're both current formats but only Stata/MP supports more than 32,767 variables. I guess that in Stata the restriction on the number of variables is enforced using the file format rather than the actual number of variables included.

From the spec for reference:

The format of .dta files has changed over time. Stata 17 writes what are known as .dta format-118 files and can read all formats of files that have ever been released. The recent history of .dta formats is

   Format    Current as of
   ---------------------------------------
     119     Stata 15 - 17 (when dataset has more than 32,767 variables)
     118     Stata 14 - 17
     117     Stata 13 
     116     internal; never released
     115     Stata 12
     114     Stata 10
     113     Stata  8
   ---------------------------------------