7
\$\begingroup\$

For a while now we've had "Clippy", which among other things has support for detecting when there's a lot of code posted but not a lot of description.

The feature is nice and comes with good intentions but a lot of users have had problems with this. They've encountered this restriction so they are unable to post, which leads to them either not posting at all, or filling up their post with more or less garbage.

I feel that the current situation is not desired. What should we do about this?

\$\endgroup\$
8
\$\begingroup\$

I dug into the implementation of this, since I couldn't recall whether fractional ratios were possible. And, good news, it looks like it's a bit more flexible than I'd remembered:

  1. If this check is enabled, ALL questions must include at least 1 character of not-code if there's even a single line of code, regardless of what the ratio is set to (so even a 0 explanation-to-code ratio will require something).
  2. The ratio itself allows fractional values - so it's possible to require only a single character of explanation for every 2 lines of code, etc.
  3. There's a maximum explanation cut-off beyond which more code doesn't require further explanation; by default, that's 500 characters of explanation, but we could reduce that.

On most sites, I wouldn't mess with this too much - the common problem on sites like Stack Overflow or Software Engineering is entirely too much code along with worthless bits of explanation to the tune of "it doesn't work".

But since the implicit question here is always "How can I make this code suck less?", there's probably less need for any explanation at all.

With this in mind, I've reduced the ratio to 4 characters per line of code, with a maximum required explanation of 120 characters - a short sentence.

\$\endgroup\$
  • \$\begingroup\$ Could you double-check the threshold? This code dump question was recently posted. \$\endgroup\$ – 200_success Jul 28 '17 at 19:53
  • \$\begingroup\$ I set the max reputation level for this (and the code formatting) check at 200, @200_success. 200 just seemed like a sensible number, a successful one even. Perhaps it should be higher? \$\endgroup\$ – Shog9 Jul 29 '17 at 3:10
2
\$\begingroup\$

According to the comments in this thread by Shog, this rule is implemented by a simple a ratio check, and it's not easy to customize it, for example to require a fixed minimum amount of text.

I think the amount of text required to understand some code doesn't grow linearly with the amount of code. It grows slower. It seems that questions with longer code have a hard time adding enough text to meet this new requirement. If this rule cannot be adapted better to longer code, I think it would be better to disable it.

\$\endgroup\$
  • 1
    \$\begingroup\$ I think that the original request was to weed out code-only posts. We don't have to doable it altogether. Any non-zero threshold (e.g. 0.3%, which is 200 bytes per 64 kiB) would be useful. \$\endgroup\$ – 200_success Jul 23 '17 at 13:27
0
\$\begingroup\$

Can we add a link in Clippy's text to Simon's Guide to a good question?

That's the link I always recommend when users have problems filling up their post with text. In addition to giving them things to write as a description it also gives suggestions on how to write a good question overall.

\$\endgroup\$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .