Print representation of structs in Emacs 26 breaks tests

[xiongtx]

Some tests are failing when run against Emacs 26 due to the new #s struct print syntax, which seems to be due to the use of records to represent structs.

[gonewest818]

I'm curious about this.

FAILED: Expected `(nrepl-bdecode "lli1ei2ei3eeli5ei6eee")' to be `equal' to 
`([cl-struct-queue nil nil] . [cl-struct-nrepl-response-queue (((1 2 3) (5 6))) (((1 2 3) (5 6))) nil])', 
but instead it was `(#s(queue nil nil) . #s(nrepl-response-queue (((1 2 3) (5 6))) (((1 2 3) (5 6))) nil))'
which does not match because: (car (different-types #s(queue nil nil) [cl-struct-queue nil nil])).

It doesn't look like the reader supports #s syntax for arbitrary structs, so how does one specify the correct match for that test?

Would you be okay with constructing the match on the fly, as in

    (it "decodes queues"
      (expect (nrepl-bdecode "lli1ei2ei3eeli5ei6eee")
              :to-equal (cons
                         (make-queue)
                         (let ((n (nrepl-response-queue)))
                           (queue-enqueue n `((1 2 3) (5 6)))
                           n))))

I don't like mixing tests (the nrepl-decode and the constructors for those structures) but it does work.

[xiongtx]

The Emacs 26 reader should support #s for structs. Is that not the case?

[gonewest818]

I may be mistaken. I saw documentation somewhere that said that #s notation was supported only for hash-tables. I didn’t understand why that would be the case, but the way it was written it seemed like they were going out of their way to make the point.

…

On Dec 15, 2017, at 12:58 AM, Tianxiang Xiong ***@***.***> wrote: The Emacs 26 reader_should_ support #s for structs. Is that not the case? — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

[xiongtx]

#s notation was supported only for hash-tables

Funny you should mention it--I actually asked about this on emacs-devel. Seems the decision by Stefan Monnier to use #s (which corresponds to #s for structures in CL) was a controversial one.

Ultimately I think Stefan made the right choice--closer alignment w/ CL is a good thing.

[gonewest818]

Ok this works in emacs 26.0.90

 (expect (nrepl-bdecode "lli1ei2ei3eeli5ei6eee")
                  :to-equal (cons
                             #s(queue nil nil)
                             #s(nrepl-response-queue (((1 2 3) (5 6))) (((1 2 3) (5 6))) nil)))))

But it’s not valid code for older versions of emacs, so presumably a macro or something is needed to conditionally compile this new test canonical only when the emacs version is 26 or greater.

[xiongtx]

Building structs on-the-fly seems to be the best way to avoid this problem.