Fix broken survey submissions

[tfausak]

This issue is to figure out how to react to this bug: 3ca5432

Of the 1,196 survey submissions that I have received so far, 763 of them suffer from this bug. That means for any of the 25 questions that allowed multiple answer choices, I only recorded a single answer choice rather than all of them. There is no way for me to recover this data. The best I can do is either reach out to these people directly and ask them to take the survey again or make some kind of big announcement and hope everyone sees it. Perhaps both.

One slight wrinkle with that plan is that of the 763 potentially broken submissions, only 345 include an email address. That means there are 418 submissions where I cannot reach the person who submitted it.

[tfausak]

The fix for this bug was deployed after 2020-11-02T12:28:09Z. Any submission received at or before that time had this problem. Any submission received after that time is fine.

[tfausak]

These are the questions that allowed multiple answer choices:

• If you do not use Haskell, why not?
• Where do you use Haskell?
• If you do not use Haskell at work, why not?
• Which programming languages other than Haskell are you fluent in?
• Which types of software do you develop with Haskell?
• Which industries do you use Haskell in?
• Which platforms do you develop Haskell on?
• Which platforms do you target?
• Which Haskell compilers do you use?
• Which installation methods do you use for your Haskell compiler?
• How has upgrading your Haskell compiler broken your code in the last year?
• Which versions of GHC do you use?
• Which build tools do you use for Haskell?
• Which editors do you use for Haskell?
• Which version control systems do you use for Haskell?
• Where do you get Haskell packages from?
• Which tools do you use to test Haskell code?
• Which tools do you use to benchmark Haskell code?
• Which tools do you use to deploy Haskell applications?
• Where do you deploy Haskell applications?
• Where do you interact with the Haskell community?
• Which of the following Haskell topics would you like to see more written about?
• Did you take any previous surveys?

[tfausak]

I'm really not sure what to do here. Even if I manage to successfully email 345 people, they still need to actually receive the email, not mark it as spam, and act on it in some reasonable time frame. And even with all that I'd still need to throw out 418 responses.

I could announce that anyone who took the survey in the first 24 hours needs to re-take it. I suspect people wouldn't be too happy about that. I know I wouldn't be thrilled about filling out the survey again. The upside here is that I wouldn't need to email hundreds of people, and I could simply discard the broken submissions.

I could do nothing, leave a note in the post about the results, and make sure this doesn't happen next year. In that case I could either not count the broken answers at all, or perhaps count them but put them in a different bucket somehow.

[tfausak]

I could do nothing, leave a note in the post about the results, and make sure this doesn't happen next year.