- Posts: 2848
- Karma: 72
- Thank you received: 557
Generate dummy results
- Posts: 9454
- Karma: 403
- Thank you received: 1627
Make a 'false' test survey and fill information randomly break surely some settings. Managing relevance can be done but for ALL other settings, it can be very difficult ...
Question is, if it is difficult because of the task or because the codebase of LS2 isn't allowing the easiest approach.
DenisChenu wrote: Make a 'false' test survey and fill information randomly break surely some settings. Managing relevance can be done but for ALL other settings, it can be very difficult ...
Since LS3.0 is already in development I wonder how the codebase will differ.
Such a testing feature is pretty common with the established suites. They are more sophisticated since you can adjusting what and at which percentage routes are choosen. But mostly you will use such a tool to check if you have produces some dead-ends with filters or messed up some conditions.
But even just filling results into the database to check for limits based on server configuration, database or code when doing certain actions will help debugging.
But I got the impression, that new features are breaking more and more the codebase of LS2.0 it might be time pause feature requests at all till LS 3.0.
Desktopbased survey tools (which generate the websurveycode) sometimes offer to simulate browser users.
holch wrote: I am not an expert on this, but coulndn't a bot fill in the form randomly? With this you should be able to fill the database and calculate the expected outcomes for each path.
E.g. like this :
On webbased survey tools I haven't seen external bots as a feature, but mostly internal routines which fill up the database, but will offer features like skip logic, use certain filter with a certain percentage and define answers. But that's a long way to go. Question is which is a approach that will help developer and users, without costing months of development time. Poweruser might want to use this kind of feature, but e.g the lacking functions of EM seem to be more important. manual.limesurvey.org/Expression_Manager..._or_Being_Considered
In general, bots are able to fill out online forms but when it comes to complex surveys, there are many, any things to watch out for. Just some examples (there are probably some dozens):
1. Using validation for allowed numeric ranges (e.g. 0 to 100 for percentages).
2. Using "sum totals X".
3. Using a text field with regex validation e.g. for valid email addresses.
4. Simulating real clicks to trigger JS based filters e.g. for array filter and firing conditions.
5. Limiting input to numeric values.
6. Limiting input to integers only.
7. Dealing with questions hidden by conditions.
Of course, such a tool would be helpful, but it will probably take weeks to develop this.
Maybe this can be a Google Summer of Code 2016 task if Google is so kind to accept Limesurvey next year.
Best regards/Beste Grüße,
Dr. Marcel Minke
Need Help? We offer professional Limesurvey support