Hi Chris, I just pushed an update to struct-plus-plus that I think does what you want. https://pkgd.racket-lang.org/pkgn/package/struct-plus-plus
The package server is telling me "This package needs documentation". I'm not sure why that is, since I've got scribblings/strut-plus-plus.scrbl and I thought that was all I needed. Regardless, here's the link to the HTML file until I can get it sorted: https://www.dropbox.com/s/qj1krqau8nonhxq/struct-plus-plus.html?dl=0 Here's how to implement what you were looking for: #lang racket (require struct-plus-plus) (struct Temp (val)) (struct Label (val)) (struct VarValue (val)) ; The requirements you gave were: ;; "In this case, if name is Temp? then global? has to be #f." ;; "If name is Label? then global? can be #t or #f." (struct++ Result ([name (or/c Temp? Label?)] [global? any/c (negate false?)] ; coerce to #t or #f [value (or/c VarValue? #f)]) (#:rule ("check name" #:check (name global?) [(or (Label? name) (and (Temp? name) (false? global?)))])) #:transparent ) ; Just for fun, I've contracted global? to accept any value but coerce ; it to #t / #f. It's effectively still a boolean field but it's more ; convenient. (Result++ #:name (Temp 7) #:global? #f #:value #f) (with-handlers ([identity identity]) (Result++ #:name (Temp 7) #:global? #t #:value #f)) (Result++ #:name (Label "foo") #:global? #f #:value #f) (Result++ #:name (Label "foo") #:global? #t #:value #f) (Result++ #:name (Label "foo") #:global? 'ok #:value #f) ; 'ok is coerced to #t Output if you run the above: (Result #<Temp> #f #f) (exn:fail:contract "check name: check failed\n name: #<Temp>\n global?: #t" #<continuation-mark-set>) (Result #<Label> #f #f) (Result #<Label> #t #f) (Result #<Label> #t #f) On Fri, Feb 15, 2019 at 7:53 AM Chris GauthierDickey <chris...@gmail.com> wrote: > > Still having trouble though, can't get struct/dc to work in the way I'd like. > I tried to simplify my example: > > (define Result/c > (struct/dc Result > (a number?) > (b boolean?) > ; in other words, if a is 0, then b must be > ; false or the contract is broken > #:inv (a b) (if (= a 0) > (eq? b #f) > #t))) > (provide (contract-out [Result Result/c])) > > The documentation on #:inv says it must return a non-#f value, which I assume > means it acts like a predicate? #f if the contract fails, something else > otherwise? In any case, this doesn't work because it says: > > Result: broke its own contract > promised: Result? > produced: #<procedure:Result> > in: (struct/dc > Result > (a number?) > (b boolean?) > #:inv > (a b) > ...) > > I also tried specifying all of the struct functions that are created with > their own contracts, like Result-a, Result-b, Result? and then using Result/c > for Result, and this fails too. > > I think I understand how this is being used in the example in the > documentation with #:lazy, but it's not quite what I'm looking for: I'm > looking to use a contract on the struct constructor (in particular, which is > why I thought #:inv would be the way to go) where one field initialization > depends on another. > > As an aside, I also tried the following: > > (provide (contract-out [Result (or/c (number? boolean? . -> . Result?) > (0 #f . -> . Result?))])) > > But the contract system complained that both might match (I mean I can see > that one is more specific, but I understand it's not a theorem prover, haha). > > Thoughts? > Chris > > On Fri, Feb 15, 2019 at 7:56 AM Chris GauthierDickey <chris...@gmail.com> > wrote: >> >> I believe it should be since bst/c is supposed to be a contract. Also I >> noticed the docs at >> https://docs.racket-lang.org/reference/data-structure-contracts.html show a >> similar example: >> >> ; bst-between : number number -> contract >> ; builds a contract for binary search trees >> ; whose values are between low and high >> (define (bst-between/c low high) >> (or/c null? >> (struct/dc node [val (between/c low high)] >> [left (val) #:lazy (bst-between/c low val)] >> [right (val) #:lazy (bst-between/c val high)]))) >> >> >> On Fri, Feb 15, 2019 at 1:31 AM David Storrs <david.sto...@gmail.com> wrote: >>> >>> On Fri, Feb 15, 2019 at 12:18 AM David Storrs <david.sto...@gmail.com> >>> wrote: >>> > >>> > On Thu, Feb 14, 2019 at 9:08 PM Robby Findler >>> > <ro...@eecs.northwestern.edu> wrote: >>> > > >>> > > This is what struct/dc is for. Let me know if the docs let you down! >>> > > >>> > > Robby >>> > >>> > This is good to hear about, because I'm actually in the middle of >>> > writing something that it will help with. Thanks, Robby. >>> > >>> > One question: Is there a typo in the example code or am I missing >>> > something? >>> > >>> > (struct bt (val left right)) >>> > (define (bst/c lo hi) >>> > (or/c #f >>> > (struct/dc bt >>> > [val (between/c lo hi)] >>> > [left (val) #:lazy (bst lo val)] ; should that be (bt >>> > lo val)? >>> > [right (val) #:lazy (bst val hi)]))) ; and this be (bt val >>> > hi) ? >>> >>> Ugh. I hate it when I make significant typos in code questions. I >>> should have asked if it was supposed to (bst/c lo val) and (bst/c val >>> hi). (i.e., is there a missing /c in the code?) If not, where is bst >>> defined? >>> >>> > >>> > Also, am I correct that the contract is not actually attached to the >>> > bt structure at this point? That's a thing that would be done >>> > elsewhere, probably in the provide statement. Yes? >>> > >>> > > >>> > > On Thu, Feb 14, 2019 at 7:57 PM Chris GauthierDickey >>> > > <chris...@gmail.com> wrote: >>> > >> >>> > >> I'm wondering if it's possible to have a contract for a struct that >>> > >> only allows certain kinds of initializations. For example, I have this: >>> > >> >>> > >> (provide (contract-out >>> > >> [struct Result ((name (or/c Temp? Label?)) (global? >>> > >> boolean?) (value (or/c VarValue? #f)))])) >>> > >> >>> > >> But it's too general. What I'd really like to have is a contract that >>> > >> combines them like an or/c (I can see what this would cause problems, >>> > >> just wondering if there's a clever way around it): >>> > >> (provide (contract-out >>> > >> [or/c (struct Result ((name Temp?) (global? #f) (value (or/c >>> > >> VarValue? #f)))) >>> > >> (struct Result ((name Label?) (global? boolean?) >>> > >> (value (or/c (VarValue #f)))))])) >>> > >> >>> > >> In this case, if name is Temp? then global? has to be #f. If name is >>> > >> Label? then global? can be #t or #f. I'd like to use struct cause it >>> > >> provides all the contracts for accessors and such. Note, I can use a >>> > >> #:guard on the struct, but wondered if it was possible to do something >>> > >> along these lines with a contract. >>> > >> >>> > >> Thanks! >>> > >> Chris >>> > >> >>> > >> >>> > >> -- >>> > >> You received this message because you are subscribed to the Google >>> > >> Groups "Racket Users" group. >>> > >> To unsubscribe from this group and stop receiving emails from it, send >>> > >> an email to racket-users+unsubscr...@googlegroups.com. >>> > >> For more options, visit https://groups.google.com/d/optout. >>> > > >>> > > -- >>> > > You received this message because you are subscribed to the Google >>> > > Groups "Racket Users" group. >>> > > To unsubscribe from this group and stop receiving emails from it, send >>> > > an email to racket-users+unsubscr...@googlegroups.com. >>> > > For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
