Nothing Special   »   [go: up one dir, main page]
Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Make Ulf Norell's define example work in all backends #363

Closed
5 tasks done
andreasabel opened this issue Apr 28, 2021 · 6 comments
Closed
5 tasks done

Make Ulf Norell's define example work in all backends #363

andreasabel opened this issue Apr 28, 2021 · 6 comments
Assignees
@andreasabel
Labels
Agda Issues of the Agda backend bug C++ C define Issue with define pragma Java lists Concerning list categories and separator/terminator/delimiter pragmas OCaml regression in master
Milestone
2.9.2

Comments

@andreasabel
Copy link
Member
andreasabel commented Apr 28, 2021 •
edited
Loading

The original example for define by Ulf Norell is currently only accepted by the Haskell backends.

bnfc/examples/DefinedRules.cf

Lines 22 to 23 in 3da1b0e

define if e s = If e s (Block []) ;
define for i c s b = Block [i, While c (Block [b, s])] ;

  • Agda: wrong list constructors
  • C: list constructors rendered wrongly as make_(:) and make_[].
  • CPP: missing sanitization of if and for (keyword clash, regression), wrong list constructors.
  • Ocaml: missing sanitization (regression), list cons (:) (a,b) instead of (a :: b).
  • Java: missing sanitization (regression), complaint that if clashes with If. Works in 2.9.1 with --force.

Missing sanitization is a regression wrt. 2.9.1, introduced by #287.
The list constructors probably never worked.
@andreasabel andreasabel added bug OCaml C++ C Java lists Concerning list categories and separator/terminator/delimiter pragmas Agda Issues of the Agda backend define Issue with define pragma regression in master labels Apr 28, 2021
@andreasabel andreasabel added this to the 2.9.2 milestone Apr 28, 2021
@andreasabel andreasabel mentioned this issue Apr 28, 2021
Closed
@shlevy
Copy link
Contributor
shlevy commented Apr 30, 2021

@andreasabel We just hit:

C: list constructors rendered wrongly as make_(:) and make_[]

Any pointers to fix?

shlevy added a commit to scarf-sh/bnfc that referenced this issue Apr 30, 2021
@shlevy
C: Properly implement defined rules using the empty list.
6b8b9e0 
Part of BNFC#363.
@shlevy shlevy mentioned this issue Apr 30, 2021
Closed
andreasabel added a commit that referenced this issue May 2, 2021
@andreasabel
[ #363 ] move examples/definedRules.cf to examples/rules/define.cf
0ae57c4
andreasabel added a commit that referenced this issue May 2, 2021
@andreasabel
[ #363 ] example text for examples/define/test.cf
630c452
@andreasabel andreasabel self-assigned this May 18, 2021
@andreasabel
Copy link
Member Author

I am now working to extend the expression type by typing information, so that the printers can pick the right list constructors.

andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 ] Pretty Exp' instead of Show Exp
742034c 
Preparing for a refactoring of the TypeChecker to retain more
information about the list constructors.
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 ] move predicates isNilFun etc. to IsFun class
c106dc3 
Preparing for an algebraic type `Label` with `instance IsFun Label`.
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 refactor ] Pretty Rule/Cat instead of Show Rule/Cat
860980e 
Show Cat is retained for Hspec, but not used in the main code base.
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 refactor ] retain typing information constructed by TypeChecker
7808875 
Typing information is preserved in Exp and the Define pragma.

Not used yet in the backends.
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 ] C: fixed define with list expressions
8eccfdc 
- NULL instead of make_[]
- make_ListFoo instead of make_(:)
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 #287 ] fixed for Java
fdbac3b 
- do not invoke the type checker a second time
- 'cons' needed in AbsynDef.java
- sanitize Java keywords clashing with definitions
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 #287 ] fixed for Ocaml: cons, sanitization of keywords
d67f07e
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 ] fixed for Agda: list and Integer literals
df4db41
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 ] fixed for Haskell/Functor
b420b9e 
List-cons does not need a type parameter.
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 #287 ] fixed for C++: cons, sanitization of keywords
2dbcb9b
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 ] test case: Norell's example for 'define'
4945079 
This test has a non-linear definition:
```
define inc x        = Assign x (EOp (EVar x) Plus (EInt 1)) ;
```
This creates syntax trees that not trees bug DAGs.
```
  make_Assign (x, make_EOp (make_EVar(x), make_Plus(), make_EInt(1)))
```
E.g. here, `x` is shared.  This causes problems for the C deallocater
which assumes that it is dealing with a tree.

This issue is left open for now.
andreasabel added a commit that referenced this issue May 18, 2021
@andreasabel
[ #363 #366 ] CHANGELOG
228e2ef
@andreasabel
Copy link
Member Author
andreasabel commented May 18, 2021 •
edited
Loading

Remaining problem (C): non-linear definitions, like:

bnfc/examples/define/test.cf

Line 25 in 228e2ef

define inc x = Assign x (EOp (EVar x) Plus (EInt 1)) ;

This creates syntax trees that not trees but DAGs.

  make_Assign (x, make_EOp (make_EVar(x), make_Plus(), make_EInt(1)))

E.g. here, x is shared. This causes problems for the C deallocator which assumes that it is dealing with a tree.

@andreasabel andreasabel mentioned this issue May 18, 2021
Open
8 tasks
@shlevy
Copy link
Contributor
shlevy commented May 18, 2021

the C deallocator

A bit off topic, but is there a deallocation function provided with the generated files? I haven't seen one, I've just been manually freeing recursively down the generated structs.

@andreasabel
Copy link
Member Author

@shlevy : The deallocation is a new, yet unreleased functionality added in a9d360d (#348).

andreasabel added a commit that referenced this issue May 19, 2021
@andreasabel
[ #363 ] C: fix doctest for prDefH
94cc080
andreasabel added a commit that referenced this issue Jun 5, 2021
@andreasabel
[ #363 ] C: clone duplicate uses in definitions to get absyn _tree_
3265143 
Sharing introduced by definitions would result in a DAG, crashing the
deallocator which assumes it is a tree.
@andreasabel
Copy link
Member Author

Remaining problem (C): non-linear definitions, like:

This has been fixed by introduction of clone_* and the cloning of duplicate components to prevent pointer aliasing and to preserve the tree-property.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@andreasabel andreasabel

Labels
Agda Issues of the Agda backend bug C++ C define Issue with define pragma Java lists Concerning list categories and separator/terminator/delimiter pragmas OCaml regression in master
Projects
None yet
Milestone
2.9.2
Development

No branches or pull requests
2 participants
@shlevy @andreasabel