Make compiler throw warnings

[VMatthijs]

We need to throw warnings for deprecated constructs.

Are there any more that are desired?

[bob-carpenter]

I think we should just throw warnings for deprecations and leave everything else to some kind of lint mode extension.

[VMatthijs]

Thanks, @bob-carpenter !

[VMatthijs]

@bob-carpenter , @seantalts , I currently have a local solution to this issue. However, it generates quite wordy warnings for a lot of files and I'm not sure what is most desirable. It generates warnings like

Warning: deprecated language construct used at file "wishart2x2.stan", line 51, column 19:
   -------------------------------------------------
    49:    //     | rho * sd2               rho * sd1        sd1 * sd2 * (1 - rho^2)   |
    50:
    51:    increment_log_prob(log(2.0 * sd1)
                            ^
    52:                       + log(2.0 * sd2)
    53:                       + log(sd1 * sd2 * (1.0 - rho * rho)));
   -------------------------------------------------

increment_log_prob(...); is deprecated and will be removed in the future. Use target += ...; instead.

for every use of a deprecated construct in a file. Some files have many deprecated constructs in them, so we get a lot of these warnings. For instance, on the file

model {
   real x;
   x <- 24.2; #wrong
   #again
   increment_log_prob(14.0);
   print(get_lp());
}

we get the warnings

Warning: deprecated language construct used at file "test2.stan", line 3, column 6:
   -------------------------------------------------
     1:  model {
     2:     real x;
     3:     x <- 24.2; #wrong
               ^
     4:     #again
     5:     increment_log_prob(14.0);
   -------------------------------------------------

assignment operator <- is deprecated in the Stan language; use = instead.


Warning: deprecated language construct used at file "test2.stan", line 3, column 14:
   -------------------------------------------------
     1:  model {
     2:     real x;
     3:     x <- 24.2; #wrong
                       ^
     4:     #again
     5:     increment_log_prob(14.0);
   -------------------------------------------------

Comments beginning with # are deprecated. Please use // in place of # for line comments.


Warning: deprecated language construct used at file "test2.stan", line 4, column 3:
   -------------------------------------------------
     2:     real x;
     3:     x <- 24.2; #wrong
     4:     #again
            ^
     5:     increment_log_prob(14.0);
     6:     print(get_lp());
   -------------------------------------------------

Comments beginning with # are deprecated. Please use // in place of # for line comments.


Warning: deprecated language construct used at file "test2.stan", line 5, column 20:
   -------------------------------------------------
     3:     x <- 24.2; #wrong
     4:     #again
     5:     increment_log_prob(14.0);
                             ^
     6:     print(get_lp());
     7:  }
   -------------------------------------------------

increment_log_prob(...); is deprecated and will be removed in the future. Use target += ...; instead.


Warning: deprecated language construct used at file "test2.stan", line 6, column 14:
   -------------------------------------------------
     4:     #again
     5:     increment_log_prob(14.0);
     6:     print(get_lp());
                       ^
     7:  }
   -------------------------------------------------

get_lp() function is deprecated. It will be removed in a future release. Use target() instead.

I could of course make the warnings less verbose or make sure at most one warning gets thrown per file, but that might also make them less useful. What is desirable here in your opinions?

[VMatthijs]

Update: I have now pushed my local implementation with the behaviour described above. If you would prefer a different kind of behaviour, I'm happy to change it!

[bob-carpenter]

I'd be OK either way---nagging people with a warning on every one or just flagging the first. The former seems like a lot less stateful and thus easier to maintain, so that seems like the natural way to go.

…

On Jan 11, 2019, at 10:18 AM, Matthijs Vákár ***@***.***> wrote: Update: I have now pushed my local implementation with the behaviour described above. If you would be prefer a different kind of behaviour, I'm happy to change it! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[VMatthijs]

Great! Then, this should be solved. Thanks, @bob-carpenter !