We are noticing differences in the SPSS data file depending on whether we extract the SPSS file for ‘All responses’ versus ‘Completed responses only’.

It appears that the length of the text answers are longer for the ‘completed responses only’ syntax file, which show full text answers, while the text answers appear to be truncated in the “All responses” SPS file.

However, the truncation does not appear to be consistent across variables. Is this programmed on purpose (i.e. because there is simply a larger volume of data for the ‘All responses’ file, longer text answers are automatically truncated), or is there an error occurring?

LS 2.67.3 build 170728

I'd say it's a bug.
Could you please a bug report and attach to it a lsa file?

I've reported as bug : bugs.limesurvey.org/view.php?id=12697

I'm willing to fund fixing this if we can expedite it, as it is causing us a major problem with our data analysis.

I added this information to the bug report, but then wondered if maybe it isn't a bug, and instead is correct behaviour?

I notice that the syntax file for Complete Only and Incomplete Only contain different values. I've attached a comparison of the relevant parts of the SPS file (that are different).

Is there an explanation for the differences?

Hi, blocka,
I'm sure it's not your desired answer, but I had the same problem. Probably (it's visible in your comparison pdf) length of text variables for "Completed responses only" is miscalculated. So, you have a few options how to handle with this.
1. Export all responses from Lime, import them to SPSS and remove all not finished (with an empty variable "Submit date").
2. Export from Lime syntax (.sps) file for "All responses" and data file (.dat) for "Completed responses"
3. If in your survey is just a few open text questions, you can edit syntax file (.sps) and change manually length of text variables (A is text variable, number behind A is a length. For example V54 A45 to V54 A200). In most of my surveys, it is the easiest way.

@Payek thanks for the input!

Yes, you are right, it is a "bug" -- and Denis Chenu @ sondages.pro has provided me with a patch, which I think is being submitted to LS core for consideration.

Yes : patch is here : github.com/LimeSurvey/LimeSurvey/pull/815 .

But still have a potential issue with 'Array (number)' question type. Please : if it's fix your issue : put it on githb (in a comment)