Reading Stata value labels fails for all value labels prior to an empty value label

[NilsEnevoldsen]

Reported as tidyverse/haven#551

Variables whose value labels appear on the label list prior to one with an empty value label are not assigned their value labels.

Step 1:

input x y z
0 0 0
end

lab def xlab 0 "Foo"
lab val x xlab

lab def ylab 0 "Bar"
lab val y ylab

lab def ylab 0 "", modify // This is one way to create an "empty" value label. It does *not* create a value label with an empty string for value 0.

lab def zlab 0 "FooBar"
lab val z zlab

label list
save example_input

Step 2: readstat example_input.dta example_output.dta

Actual example_output.dta :

Expected example_output.dta :

(Minor detail: the blueness of the 0 indicates the presence of an [empty] value label for variable y; I have no strong opinion on whether this value label should be preserved in the output file.)

[evanmiller]

Can you please provide a copy of example_input.dta? Thanks

[NilsEnevoldsen]

Yes, I have attached four example_inputs. I used Stata 14 to generate them with the following code:

save example_input_14
saveold example_input_13, version(13)
saveold example_input_12, version(12)
saveold example_input_11, version(11)

example_input_14.dta (version 118) and example_input_13.dta (version 117) exhibit the incorrect behavior when processed by ReadStat. example_input_12.dta (version 115) and example_input_11.dta (also version 115? I'm unclear on this) exhibit correct behavior when processed by ReadStat. (In the latter cases, they do both drop ylab, which I don't feel strongly about.)

example_input.zip

[evanmiller]

Thanks. See if this commit fixes things:

263ba39

[NilsEnevoldsen]

I get an error on every DTA file with 1.1.5-rc0.

% readstat example_input_16.dta example_output_16.dta
Error beginning row: Unable to write data
Converted 1 variables and 1 rows in 0.00 seconds
Error processing example_input_16.dta: The parsing was aborted (callback returned non-zero value)

[evanmiller]

Please try the dev branch rather than the RC

[NilsEnevoldsen]

Ah, right, I just saw 93e33e2

[NilsEnevoldsen]

Yes, this fix works for me, with no immediately apparent side effects. I tested example DTAs from Stata versions 11 through 16, as well as my own "in the wild" problematic example, and that reported at tidyverse/haven#551.

Thank you for your prompt fix! Will this make it into 1.1.5?

[evanmiller]

Will this make it into 1.1.5?

Yes