Fix field accessors not showing up in view in some cases

[runarorama]

For some types with lots of field accessors, they show up without field accessors when printed with edit or view, like this:

  type Mail
    = Mail
        [Personalization]
        MailAddress
        (Optional MailAddress)
        [MailAddress]
        Text
        [MailContent]
        (Optional [MailAttachment])
        (Optional Text)
        (Optional Json)
        (Optional (Map Text Text))
        (Optional [Text])
        (Optional Json)
        (Optional Instant)
        (Optional Text)
        (Optional Asm)
        (Optional Text)
        (Optional MailSettings)
        (Optional TrackingSettings)

This is caused by hashFieldAccessors returning Nothing for the first .set accessor. This is in turn caused by a type error in Typechecker.synthesize for that accessor. The note from the typechecker is:

The 2nd argument to `Mail.Mail`

          has type:  lib.base.Optional lib.base.time.Instant
    but I expected:  MailAddress

The term that's being typechecked looks like this:

(λ (User "_18". (λ (User "mail". (case Var User "mail" of [MatchCase {matchPattern = Constructor ReferenceDerived (Id "9mq7l0pr6foeuis6jujpo7hqcbsm5120lhu2nf70tccdtpuosmdfhr81foq9okme97b8kgd2h7k14pceu0nv9r2qjll416fiu5t06cg" 0) 0 [Unbound,Var,Var,Var,Var,Var,Var,Var,Var,Var,Var,Var,Var,Var,Var,Var,Var,Var], matchGuard = Nothing, matchBody = (User "_"-1. (User "_"-2. (User "_"-3. (User "_"-4. (User "_"-5. (User "_"-6. (User "_"-7. (User "_"-8. (User "_"-9. (User "_"-10. (User "_"-11. (User "_"-12. (User "_"-13. (User "_"-14. (User "_"-15. (User "_"-16. (User "_"-17. ConReferenceDerived (Id "9mq7l0pr6foeuis6jujpo7hqcbsm5120lhu2nf70tccdtpuosmdfhr81foq9okme97b8kgd2h7k14pceu0nv9r2qjll416fiu5t06cg" 0)#0 (Var User "_18") (Var User "_"-1) (Var User "_"-2) (Var User "_"-3) (Var User "_"-4) (Var User "_"-5) (Var User "_"-6) (Var User "_"-7) (Var User "_"-8) (Var User "_"-9) (Var User "_"-10) (Var User "_"-11) (Var User "_"-12) (Var User "_"-13) (Var User "_"-14) (Var User "_"-15) (Var User "_"-16) (Var User "_"-17))))))))))))))))))}])))))

This fix adds an ordinal to the variable names of the fields, so freshen should in most cases be a no-op.

After this fix the type above is rendered as:

  type Mail
    = { personalizations : [Personalization],
        from : MailAddress,
        replyTo : Optional MailAddress,
        replyToList : [MailAddress],
        subject : Text,
        content : [MailContent],
        attachments : Optional [MailAttachment],
        templateId : Optional Text,
        sections : Optional Json,
        headers : Optional (Map Text Text),
        categories : Optional [Text],
        customArgs : Optional Json,
        sendAt : Optional Instant,
        batchId : Optional Text,
        asm : Optional Asm,
        ipPoolName : Optional Text,
        mailSettings : Optional MailSettings,
        trackingSettings : Optional TrackingSettings }

[ChrisPenner]

Sorry can you explain in simpler terms what the problem was here? Were there conflicting type var names that caused it to fail to type-check, and if that's the case, is freshIn broken or was it just not scoped properly?

Great find either way BTW 🙌🏼

[aryairani]

Sorry can you explain in simpler terms what the problem was here? Were there conflicting type var names that caused it to fail to type-check, and if that's the case, is freshIn broken or was it just not scoped properly?

Great find either way BTW 🙌🏼

@ChrisPenner Our guess is that, for either misguided or arbitrary reasons, the term printer was not printing same-named-but-differently-fresh variables with distinct names, apart from pattern vars. This PR sidesteps that issue by giving these variables differently-named vars, instead of same-named-but-differently-fresh vars.

• And then we also want to go back eventually and check term printing of vars / figure out why #4752 was necessary #4765