2009-05-15 23:44:14 +02:00
|
|
|
|
;;; TREE-IL -> GLIL compiler
|
|
|
|
|
|
|
2010-01-08 12:02:00 +01:00
|
|
|
|
;; Copyright (C) 2001,2008,2009,2010 Free Software Foundation, Inc.
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-06-17 00:22:09 +01:00
|
|
|
|
;;;; This library is free software; you can redistribute it and/or
|
|
|
|
|
|
;;;; modify it under the terms of the GNU Lesser General Public
|
|
|
|
|
|
;;;; License as published by the Free Software Foundation; either
|
|
|
|
|
|
;;;; version 3 of the License, or (at your option) any later version.
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;;;;
|
2009-06-17 00:22:09 +01:00
|
|
|
|
;;;; This library is distributed in the hope that it will be useful,
|
|
|
|
|
|
;;;; but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
|
|
;;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
|
|
;;;; Lesser General Public License for more details.
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;;;;
|
2009-06-17 00:22:09 +01:00
|
|
|
|
;;;; You should have received a copy of the GNU Lesser General Public
|
|
|
|
|
|
;;;; License along with this library; if not, write to the Free Software
|
|
|
|
|
|
;;;; Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
|
|
|
|
|
;;; Code:
|
|
|
|
|
|
|
|
|
|
|
|
(define-module (language tree-il analyze)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
#:use-module (srfi srfi-1)
|
2009-07-31 00:42:58 +02:00
|
|
|
|
#:use-module (srfi srfi-9)
|
2010-01-11 01:19:16 +01:00
|
|
|
|
#:use-module (srfi srfi-11)
|
2010-02-02 23:59:03 +01:00
|
|
|
|
#:use-module (ice-9 vlist)
|
2009-05-15 23:44:14 +02:00
|
|
|
|
#:use-module (system base syntax)
|
2009-07-31 00:42:58 +02:00
|
|
|
|
#:use-module (system base message)
|
2009-11-08 01:02:08 +01:00
|
|
|
|
#:use-module (system vm program)
|
2009-05-15 23:44:14 +02:00
|
|
|
|
#:use-module (language tree-il)
|
2009-11-08 17:53:14 +01:00
|
|
|
|
#:use-module (system base pmatch)
|
2009-07-31 00:42:58 +02:00
|
|
|
|
#:export (analyze-lexicals
|
2009-11-06 10:42:45 +01:00
|
|
|
|
analyze-tree
|
|
|
|
|
|
unused-variable-analysis
|
2010-01-11 01:19:16 +01:00
|
|
|
|
unused-toplevel-analysis
|
2009-11-07 18:32:26 +01:00
|
|
|
|
unbound-variable-analysis
|
|
|
|
|
|
arity-analysis))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; Allocation is the process of assigning storage locations for lexical
|
|
|
|
|
|
;; variables. A lexical variable has a distinct "address", or storage
|
|
|
|
|
|
;; location, for each procedure in which it is referenced.
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; A variable is "local", i.e., allocated on the stack, if it is
|
|
|
|
|
|
;; referenced from within the procedure that defined it. Otherwise it is
|
|
|
|
|
|
;; a "closure" variable. For example:
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; (lambda (a) a) ; a will be local
|
|
|
|
|
|
;; `a' is local to the procedure.
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; (lambda (a) (lambda () a))
|
|
|
|
|
|
;; `a' is local to the outer procedure, but a closure variable with
|
|
|
|
|
|
;; respect to the inner procedure.
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; If a variable is ever assigned, it needs to be heap-allocated
|
|
|
|
|
|
;; ("boxed"). This is so that closures and continuations capture the
|
|
|
|
|
|
;; variable's identity, not just one of the values it may have over the
|
|
|
|
|
|
;; course of program execution. If the variable is never assigned, there
|
|
|
|
|
|
;; is no distinction between value and identity, so closing over its
|
|
|
|
|
|
;; identity (whether through closures or continuations) can make a copy
|
|
|
|
|
|
;; of its value instead.
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; Local variables are stored on the stack within a procedure's call
|
|
|
|
|
|
;; frame. Their index into the stack is determined from their linear
|
|
|
|
|
|
;; postion within a procedure's binding path:
|
2009-05-15 23:44:14 +02:00
|
|
|
|
;; (let (0 1)
|
|
|
|
|
|
;; (let (2 3) ...)
|
|
|
|
|
|
;; (let (2) ...))
|
|
|
|
|
|
;; (let (2 3 4) ...))
|
|
|
|
|
|
;; etc.
|
|
|
|
|
|
;;
|
2009-05-20 12:46:23 +02:00
|
|
|
|
;; This algorithm has the problem that variables are only allocated
|
|
|
|
|
|
;; indices at the end of the binding path. If variables bound early in
|
|
|
|
|
|
;; the path are not used in later portions of the path, their indices
|
|
|
|
|
|
;; will not be recycled. This problem is particularly egregious in the
|
|
|
|
|
|
;; expansion of `or':
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; (or x y z)
|
|
|
|
|
|
;; -> (let ((a x)) (if a a (let ((b y)) (if b b z))))
|
|
|
|
|
|
;;
|
2009-12-11 12:00:27 +01:00
|
|
|
|
;; As you can see, the `a' binding is only used in the ephemeral
|
|
|
|
|
|
;; `consequent' clause of the first `if', but its index would be
|
|
|
|
|
|
;; reserved for the whole of the `or' expansion. So we have a hack for
|
|
|
|
|
|
;; this specific case. A proper solution would be some sort of liveness
|
|
|
|
|
|
;; analysis, and not our linear allocation algorithm.
|
2009-05-20 12:46:23 +02:00
|
|
|
|
;;
|
2010-01-30 15:49:50 +01:00
|
|
|
|
;; Closure variables are captured when a closure is created, and stored in a
|
|
|
|
|
|
;; vector inline to the closure object itself. Each closure variable has a
|
|
|
|
|
|
;; unique index into that vector.
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;;
|
2009-08-07 15:35:53 +02:00
|
|
|
|
;; There is one more complication. Procedures bound by <fix> may, in
|
|
|
|
|
|
;; some cases, be rendered inline to their parent procedure. That is to
|
|
|
|
|
|
;; say,
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; (letrec ((lp (lambda () (lp)))) (lp))
|
|
|
|
|
|
;; => (fix ((lp (lambda () (lp)))) (lp))
|
|
|
|
|
|
;; => goto FIX-BODY; LP: goto LP; FIX-BODY: goto LP;
|
|
|
|
|
|
;; ^ jump over the loop ^ the fixpoint lp ^ starting off the loop
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; The upshot is that we don't have to allocate any space for the `lp'
|
|
|
|
|
|
;; closure at all, as it can be rendered inline as a loop. So there is
|
|
|
|
|
|
;; another kind of allocation, "label allocation", in which the
|
|
|
|
|
|
;; procedure is simply a label, placed at the start of the lambda body.
|
|
|
|
|
|
;; The label is the gensym under which the lambda expression is bound.
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; The analyzer checks to see that the label is called with the correct
|
|
|
|
|
|
;; number of arguments. Calls to labels compile to rename + goto.
|
|
|
|
|
|
;; Lambda, the ultimate goto!
|
|
|
|
|
|
;;
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;;
|
|
|
|
|
|
;; The return value of `analyze-lexicals' is a hash table, the
|
|
|
|
|
|
;; "allocation".
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; The allocation maps gensyms -- recall that each lexically bound
|
|
|
|
|
|
;; variable has a unique gensym -- to storage locations ("addresses").
|
|
|
|
|
|
;; Since one gensym may have many storage locations, if it is referenced
|
|
|
|
|
|
;; in many procedures, it is a two-level map.
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; The allocation also stored information on how many local variables
|
2009-08-07 15:35:53 +02:00
|
|
|
|
;; need to be allocated for each procedure, lexicals that have been
|
|
|
|
|
|
;; translated into labels, and information on what free variables to
|
|
|
|
|
|
;; capture from its lexical parent procedure.
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;;
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
;; In addition, we have a conflation: while we're traversing the code,
|
|
|
|
|
|
;; recording information to pass to the compiler, we take the
|
|
|
|
|
|
;; opportunity to generate labels for each lambda-case clause, so that
|
|
|
|
|
|
;; generated code can skip argument checks at runtime if they match at
|
|
|
|
|
|
;; compile-time.
|
|
|
|
|
|
;;
|
2010-01-30 15:49:50 +01:00
|
|
|
|
;; Also, while we're a-traversing and an-allocating, we check prompt
|
|
|
|
|
|
;; handlers to see if the "continuation" argument is used. If not, we
|
|
|
|
|
|
;; mark the prompt as being "escape-only". This allows us to implement
|
|
|
|
|
|
;; `catch' and `throw' using `prompt' and `control', but without causing
|
|
|
|
|
|
;; a continuation to be reified. Heh heh.
|
|
|
|
|
|
;;
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; That is:
|
|
|
|
|
|
;;
|
|
|
|
|
|
;; sym -> {lambda -> address}
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
;; lambda -> (labels . free-locs)
|
|
|
|
|
|
;; lambda-case -> (gensym . nlocs)
|
2010-01-30 15:49:50 +01:00
|
|
|
|
;; prompt -> escape-only?
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;;
|
2009-08-07 15:35:53 +02:00
|
|
|
|
;; address ::= (local? boxed? . index)
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
;; labels ::= ((sym . lambda) ...)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; free-locs ::= ((sym0 . address0) (sym1 . address1) ...)
|
|
|
|
|
|
;; free variable addresses are relative to parent proc.
|
|
|
|
|
|
|
|
|
|
|
|
(define (make-hashq k v)
|
|
|
|
|
|
(let ((res (make-hash-table)))
|
|
|
|
|
|
(hashq-set! res k v)
|
|
|
|
|
|
res))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
|
|
|
|
|
(define (analyze-lexicals x)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; bound-vars: lambda -> (sym ...)
|
|
|
|
|
|
;; all identifiers bound within a lambda
|
2009-08-07 15:35:53 +02:00
|
|
|
|
(define bound-vars (make-hash-table))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; free-vars: lambda -> (sym ...)
|
|
|
|
|
|
;; all identifiers referenced in a lambda, but not bound
|
|
|
|
|
|
;; NB, this includes identifiers referenced by contained lambdas
|
2009-08-07 15:35:53 +02:00
|
|
|
|
(define free-vars (make-hash-table))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; assigned: sym -> #t
|
|
|
|
|
|
;; variables that are assigned
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(define assigned (make-hash-table))
|
2009-05-20 12:46:23 +02:00
|
|
|
|
;; refcounts: sym -> count
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; allows us to detect the or-expansion in O(1) time
|
2009-08-07 15:35:53 +02:00
|
|
|
|
(define refcounts (make-hash-table))
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
;; labels: sym -> lambda
|
2009-08-07 15:35:53 +02:00
|
|
|
|
;; for determining if fixed-point procedures can be rendered as
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
;; labels.
|
2009-08-07 15:35:53 +02:00
|
|
|
|
(define labels (make-hash-table))
|
|
|
|
|
|
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; returns variables referenced in expr
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(define (analyze! x proc labels-in-proc tail? tail-call-args)
|
|
|
|
|
|
(define (step y) (analyze! y proc labels-in-proc #f #f))
|
|
|
|
|
|
(define (step-tail y) (analyze! y proc labels-in-proc tail? #f))
|
|
|
|
|
|
(define (step-tail-call y args) (analyze! y proc labels-in-proc #f
|
|
|
|
|
|
(and tail? args)))
|
|
|
|
|
|
(define (recur/labels x new-proc labels)
|
|
|
|
|
|
(analyze! x new-proc (append labels labels-in-proc) #t #f))
|
|
|
|
|
|
(define (recur x new-proc) (analyze! x new-proc '() tail? #f))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
(record-case x
|
|
|
|
|
|
((<application> proc args)
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(apply lset-union eq? (step-tail-call proc args)
|
|
|
|
|
|
(map step args)))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-12-11 12:00:27 +01:00
|
|
|
|
((<conditional> test consequent alternate)
|
|
|
|
|
|
(lset-union eq? (step test) (step-tail consequent) (step-tail alternate)))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<lexical-ref> gensym)
|
2009-05-20 12:46:23 +02:00
|
|
|
|
(hashq-set! refcounts gensym (1+ (hashq-ref refcounts gensym 0)))
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(if (not (and tail-call-args
|
|
|
|
|
|
(memq gensym labels-in-proc)
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
(let ((p (hashq-ref labels gensym)))
|
|
|
|
|
|
(and p
|
|
|
|
|
|
(let lp ((c (lambda-body p)))
|
|
|
|
|
|
(and c (lambda-case? c)
|
|
|
|
|
|
(or
|
|
|
|
|
|
;; for now prohibit optional &
|
|
|
|
|
|
;; keyword arguments; can relax this
|
|
|
|
|
|
;; restriction later
|
|
|
|
|
|
(and (= (length (lambda-case-req c))
|
|
|
|
|
|
(length tail-call-args))
|
|
|
|
|
|
(not (lambda-case-opt c))
|
|
|
|
|
|
(not (lambda-case-kw c))
|
Revert "implement #:predicate" and remove predicate from <lambda-case>
Turns out this was not a very useful idea, and semantically tricky to
boot.
This reverts commit 24bf130fd15afbc8b3a2ccdc50a027f9b6c9e623, and makes
the following additional changes:
* module/ice-9/optargs.scm (parse-lambda-case, let-optional)
(let-optional*, let-keywords, let-keywords*):
* module/language/tree-il.scm: (<lambda-case>, parse-tree-il)
(unparse-tree-il, tree-il->scheme, tree-il-fold,
make-tree-il-folder)
(post-order!, pre-order!):
* module/language/tree-il/analyze.scm (analyze-lexicals):
* module/language/tree-il/compile-glil.scm (compile-glil):
* module/language/tree-il/inline.scm (inline!): Remove all traces of
#:predicate from tree-il.
* module/ice-9/psyntax.scm (build-simple-lambda, build-lambda-case)
(chi-lambda-case): Adapt to tree-il change.
* module/ice-9/psyntax-pp.scm: Regenerated.
* module/language/brainfuck/compile-tree-il.scm (compile-body):
* module/language/ecmascript/compile-tree-il.scm (comp, comp-body):
* test-suite/tests/tree-il.test: Adapt to tree-il change.
* doc/ref/api-procedures.texi (Case-lambda): Remove mention of
#:predicate.
2009-11-05 10:22:01 +01:00
|
|
|
|
(not (lambda-case-rest c)))
|
2009-12-11 11:49:14 +01:00
|
|
|
|
(lp (lambda-case-alternate c)))))))))
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(hashq-set! labels gensym #f))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(list gensym))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<lexical-set> gensym exp)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(hashq-set! assigned gensym #t)
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(hashq-set! labels gensym #f)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(lset-adjoin eq? (step exp) gensym))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<module-set> exp)
|
2009-05-15 23:44:14 +02:00
|
|
|
|
(step exp))
|
|
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<toplevel-set> exp)
|
2009-05-15 23:44:14 +02:00
|
|
|
|
(step exp))
|
|
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<toplevel-define> exp)
|
2009-05-15 23:44:14 +02:00
|
|
|
|
(step exp))
|
|
|
|
|
|
|
|
|
|
|
|
((<sequence> exps)
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(let lp ((exps exps) (ret '()))
|
|
|
|
|
|
(cond ((null? exps) '())
|
|
|
|
|
|
((null? (cdr exps))
|
|
|
|
|
|
(lset-union eq? ret (step-tail (car exps))))
|
|
|
|
|
|
(else
|
|
|
|
|
|
(lp (cdr exps) (lset-union eq? ret (step (car exps))))))))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
((<lambda> body)
|
|
|
|
|
|
;; order is important here
|
|
|
|
|
|
(hashq-set! bound-vars x '())
|
|
|
|
|
|
(let ((free (recur body x)))
|
|
|
|
|
|
(hashq-set! bound-vars x (reverse! (hashq-ref bound-vars x)))
|
|
|
|
|
|
(hashq-set! free-vars x free)
|
|
|
|
|
|
free))
|
|
|
|
|
|
|
2009-12-11 11:49:14 +01:00
|
|
|
|
((<lambda-case> opt kw inits vars body alternate)
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
(hashq-set! bound-vars proc
|
|
|
|
|
|
(append (reverse vars) (hashq-ref bound-vars proc)))
|
|
|
|
|
|
(lset-union
|
|
|
|
|
|
eq?
|
|
|
|
|
|
(lset-difference eq?
|
2009-10-23 15:51:00 +02:00
|
|
|
|
(lset-union eq?
|
|
|
|
|
|
(apply lset-union eq? (map step inits))
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
(step-tail body))
|
|
|
|
|
|
vars)
|
2009-12-11 11:49:14 +01:00
|
|
|
|
(if alternate (step-tail alternate) '())))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
|
call-with-values can make fewer closures
* module/language/tree-il.scm: Rename let-exp and letrec-exp to let-body
and letrec-body. Add <let-values>, a one-expression let-values that
should avoid the needless creation of two closures in many common
multiple-value cases. We'll need to add an optimization pass to the
compiler to produce this form, though, as well as rewriting lambdas
into lets, etc.
I added this form instead of adding more special cases to the
call-with-values compile code because it's a useful intermediate form
-- it will allow the optimizer to perform constant folding across more
code.
* module/language/tree-il.scm (parse-tree-il, unparse-tree-il)
(tree-il->scheme, post-order!, pre-order!): Adapt to let/letrec body
renaming, and let-values.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for
renaming, and add cases for let-values.
* module/language/tree-il/compile-glil.scm (flatten): Add a new context,
`vals', used by let-values code for the values producer. Code that
produces multiple values can then jump to the let-values MV return
address directly, instead of trampolining into a procedure. Add code to
compile let-values.
2009-06-08 00:38:49 +02:00
|
|
|
|
((<let> vars vals body)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(hashq-set! bound-vars proc
|
|
|
|
|
|
(append (reverse vars) (hashq-ref bound-vars proc)))
|
|
|
|
|
|
(lset-difference eq?
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(apply lset-union eq? (step-tail body) (map step vals))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
vars))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
call-with-values can make fewer closures
* module/language/tree-il.scm: Rename let-exp and letrec-exp to let-body
and letrec-body. Add <let-values>, a one-expression let-values that
should avoid the needless creation of two closures in many common
multiple-value cases. We'll need to add an optimization pass to the
compiler to produce this form, though, as well as rewriting lambdas
into lets, etc.
I added this form instead of adding more special cases to the
call-with-values compile code because it's a useful intermediate form
-- it will allow the optimizer to perform constant folding across more
code.
* module/language/tree-il.scm (parse-tree-il, unparse-tree-il)
(tree-il->scheme, post-order!, pre-order!): Adapt to let/letrec body
renaming, and let-values.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for
renaming, and add cases for let-values.
* module/language/tree-il/compile-glil.scm (flatten): Add a new context,
`vals', used by let-values code for the values producer. Code that
produces multiple values can then jump to the let-values MV return
address directly, instead of trampolining into a procedure. Add code to
compile let-values.
2009-06-08 00:38:49 +02:00
|
|
|
|
((<letrec> vars vals body)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(hashq-set! bound-vars proc
|
|
|
|
|
|
(append (reverse vars) (hashq-ref bound-vars proc)))
|
|
|
|
|
|
(for-each (lambda (sym) (hashq-set! assigned sym #t)) vars)
|
|
|
|
|
|
(lset-difference eq?
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(apply lset-union eq? (step-tail body) (map step vals))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
vars))
|
|
|
|
|
|
|
add <fix> tree-il construct, and compile it
* libguile/vm-i-system.c (fix-closure): New instruction, for wiring
together fixpoint procedures.
* module/Makefile.am (TREE_IL_LANG_SOURCES): Add fix-letrec.scm.
* module/language/glil/compile-assembly.scm (glil->assembly): Reindent
the <glil-lexical> case, and handle 'fix for locally-bound vars.
* module/language/tree-il.scm (<fix>): Add the <fix> tree-il type and
accessors, for fixed-point bindings. This IL construct is taken from
the Waddell paper.
(parse-tree-il, unparse-tree-il, tree-il->scheme, tree-il-fold)
(pre-order!, post-order!): Update for <fix>.
* module/language/tree-il/analyze.scm (analyze-lexicals): Update for
<fix>. The difference here is that the bindings may not be assigned,
and are not marked as such. They are not boxed.
(report-unused-variables): Update for <fix>.
* module/language/tree-il/compile-glil.scm (flatten): Compile <fix> to
GLIL.
* module/language/tree-il/fix-letrec.scm: A stub implementation of
fixing letrec -- will flesh out in a separate commit.
* module/language/tree-il/inline.scm: Fix license, it was mistakenly
added with LGPL v2.1+.
* module/language/tree-il/optimize.scm (optimize!): Run the fix-letrec!
pass.
2009-08-05 17:51:40 +02:00
|
|
|
|
((<fix> vars vals body)
|
2009-08-07 19:06:15 +02:00
|
|
|
|
;; Try to allocate these procedures as labels.
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
(for-each (lambda (sym val) (hashq-set! labels sym val))
|
2009-08-07 19:06:15 +02:00
|
|
|
|
vars vals)
|
add <fix> tree-il construct, and compile it
* libguile/vm-i-system.c (fix-closure): New instruction, for wiring
together fixpoint procedures.
* module/Makefile.am (TREE_IL_LANG_SOURCES): Add fix-letrec.scm.
* module/language/glil/compile-assembly.scm (glil->assembly): Reindent
the <glil-lexical> case, and handle 'fix for locally-bound vars.
* module/language/tree-il.scm (<fix>): Add the <fix> tree-il type and
accessors, for fixed-point bindings. This IL construct is taken from
the Waddell paper.
(parse-tree-il, unparse-tree-il, tree-il->scheme, tree-il-fold)
(pre-order!, post-order!): Update for <fix>.
* module/language/tree-il/analyze.scm (analyze-lexicals): Update for
<fix>. The difference here is that the bindings may not be assigned,
and are not marked as such. They are not boxed.
(report-unused-variables): Update for <fix>.
* module/language/tree-il/compile-glil.scm (flatten): Compile <fix> to
GLIL.
* module/language/tree-il/fix-letrec.scm: A stub implementation of
fixing letrec -- will flesh out in a separate commit.
* module/language/tree-il/inline.scm: Fix license, it was mistakenly
added with LGPL v2.1+.
* module/language/tree-il/optimize.scm (optimize!): Run the fix-letrec!
pass.
2009-08-05 17:51:40 +02:00
|
|
|
|
(hashq-set! bound-vars proc
|
|
|
|
|
|
(append (reverse vars) (hashq-ref bound-vars proc)))
|
2009-08-07 19:06:15 +02:00
|
|
|
|
;; Step into subexpressions.
|
|
|
|
|
|
(let* ((var-refs
|
|
|
|
|
|
(map
|
|
|
|
|
|
;; Since we're trying to label-allocate the lambda,
|
|
|
|
|
|
;; pretend it's not a closure, and just recurse into its
|
|
|
|
|
|
;; body directly. (Otherwise, recursing on a closure
|
|
|
|
|
|
;; that references one of the fix's bound vars would
|
|
|
|
|
|
;; prevent label allocation.)
|
|
|
|
|
|
(lambda (x)
|
|
|
|
|
|
(record-case x
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
((<lambda> body)
|
|
|
|
|
|
;; just like the closure case, except here we use
|
|
|
|
|
|
;; recur/labels instead of recur
|
|
|
|
|
|
(hashq-set! bound-vars x '())
|
|
|
|
|
|
(let ((free (recur/labels body x vars)))
|
|
|
|
|
|
(hashq-set! bound-vars x (reverse! (hashq-ref bound-vars x)))
|
|
|
|
|
|
(hashq-set! free-vars x free)
|
|
|
|
|
|
free))))
|
2009-08-07 19:06:15 +02:00
|
|
|
|
vals))
|
|
|
|
|
|
(vars-with-refs (map cons vars var-refs))
|
|
|
|
|
|
(body-refs (recur/labels body proc vars)))
|
|
|
|
|
|
(define (delabel-dependents! sym)
|
|
|
|
|
|
(let ((refs (assq-ref vars-with-refs sym)))
|
|
|
|
|
|
(if refs
|
|
|
|
|
|
(for-each (lambda (sym)
|
|
|
|
|
|
(if (hashq-ref labels sym)
|
|
|
|
|
|
(begin
|
|
|
|
|
|
(hashq-set! labels sym #f)
|
|
|
|
|
|
(delabel-dependents! sym))))
|
|
|
|
|
|
refs))))
|
|
|
|
|
|
;; Stepping into the lambdas and the body might have made some
|
|
|
|
|
|
;; procedures not label-allocatable -- which might have
|
|
|
|
|
|
;; knock-on effects. For example:
|
|
|
|
|
|
;; (fix ((a (lambda () (b)))
|
|
|
|
|
|
;; (b (lambda () a)))
|
|
|
|
|
|
;; (a))
|
|
|
|
|
|
;; As far as `a' is concerned, both `a' and `b' are
|
|
|
|
|
|
;; label-allocatable. But `b' references `a' not in a proc-tail
|
|
|
|
|
|
;; position, which makes `a' not label-allocatable. The
|
|
|
|
|
|
;; knock-on effect is that, when back-propagating this
|
|
|
|
|
|
;; information to `a', `b' will also become not
|
|
|
|
|
|
;; label-allocatable, as it is referenced within `a', which is
|
|
|
|
|
|
;; allocated as a closure. This is a transitive relationship.
|
|
|
|
|
|
(for-each (lambda (sym)
|
|
|
|
|
|
(if (not (hashq-ref labels sym))
|
|
|
|
|
|
(delabel-dependents! sym)))
|
|
|
|
|
|
vars)
|
|
|
|
|
|
;; Now lift bound variables with label-allocated lambdas to the
|
|
|
|
|
|
;; parent procedure.
|
|
|
|
|
|
(for-each
|
|
|
|
|
|
(lambda (sym val)
|
|
|
|
|
|
(if (hashq-ref labels sym)
|
|
|
|
|
|
;; Remove traces of the label-bound lambda. The free
|
|
|
|
|
|
;; vars will propagate up via the return val.
|
|
|
|
|
|
(begin
|
|
|
|
|
|
(hashq-set! bound-vars proc
|
|
|
|
|
|
(append (hashq-ref bound-vars val)
|
|
|
|
|
|
(hashq-ref bound-vars proc)))
|
|
|
|
|
|
(hashq-remove! bound-vars val)
|
|
|
|
|
|
(hashq-remove! free-vars val))))
|
|
|
|
|
|
vars vals)
|
|
|
|
|
|
(lset-difference eq?
|
|
|
|
|
|
(apply lset-union eq? body-refs var-refs)
|
|
|
|
|
|
vars)))
|
add <fix> tree-il construct, and compile it
* libguile/vm-i-system.c (fix-closure): New instruction, for wiring
together fixpoint procedures.
* module/Makefile.am (TREE_IL_LANG_SOURCES): Add fix-letrec.scm.
* module/language/glil/compile-assembly.scm (glil->assembly): Reindent
the <glil-lexical> case, and handle 'fix for locally-bound vars.
* module/language/tree-il.scm (<fix>): Add the <fix> tree-il type and
accessors, for fixed-point bindings. This IL construct is taken from
the Waddell paper.
(parse-tree-il, unparse-tree-il, tree-il->scheme, tree-il-fold)
(pre-order!, post-order!): Update for <fix>.
* module/language/tree-il/analyze.scm (analyze-lexicals): Update for
<fix>. The difference here is that the bindings may not be assigned,
and are not marked as such. They are not boxed.
(report-unused-variables): Update for <fix>.
* module/language/tree-il/compile-glil.scm (flatten): Compile <fix> to
GLIL.
* module/language/tree-il/fix-letrec.scm: A stub implementation of
fixing letrec -- will flesh out in a separate commit.
* module/language/tree-il/inline.scm: Fix license, it was mistakenly
added with LGPL v2.1+.
* module/language/tree-il/optimize.scm (optimize!): Run the fix-letrec!
pass.
2009-08-05 17:51:40 +02:00
|
|
|
|
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
((<let-values> exp body)
|
|
|
|
|
|
(lset-union eq? (step exp) (step body)))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
|
2010-02-18 17:23:49 +01:00
|
|
|
|
((<dynwind> body winder unwinder)
|
2010-01-30 15:49:50 +01:00
|
|
|
|
(lset-union eq? (step body) (step winder) (step unwinder)))
|
|
|
|
|
|
|
add dynlet to tree-il
* module/language/tree-il.scm (<dynlet>, dynlet?, make-dynlet)
(dynlet-src, dynlet-fluids, dynlet-vals, dynlet-body): New tree-il
construct, mapping to `with-fluids'.
(parse-tree-il, unparse-tree-il, tree-il->scheme, tree-il-fold):
(make-tree-il-folder, post-order!, pre-order!): Wire it up.
* module/language/tree-il/analyze.scm (analyze-lexicals): Add dynlet
support.
2010-02-18 18:05:24 +01:00
|
|
|
|
((<dynlet> fluids vals body)
|
|
|
|
|
|
(apply lset-union eq? (step body) (map step (append fluids vals))))
|
|
|
|
|
|
|
2010-02-19 11:42:00 +01:00
|
|
|
|
((<dynref> fluid)
|
|
|
|
|
|
(step fluid))
|
|
|
|
|
|
|
|
|
|
|
|
((<dynset> fluid exp)
|
|
|
|
|
|
(lset-union eq? (step fluid) (step exp)))
|
|
|
|
|
|
|
2010-02-18 23:56:12 +01:00
|
|
|
|
((<prompt> tag body handler)
|
2010-02-22 20:19:32 +01:00
|
|
|
|
(lset-union eq? (step tag) (step body) (step handler)))
|
2010-01-30 15:49:50 +01:00
|
|
|
|
|
abort always dispatches to VM bytecode, to detect same-invocation aborts
* libguile/control.h:
* libguile/control.c (scm_c_make_prompt): Take an extra arg, a cookie.
Continuations will be rewindable only if the abort has the same cookie
as the prompt.
(scm_at_abort): Redefine from scm_abort, and instead of taking rest
args, take the abort values as a list directly. Also, don't allow
rewinding, because we won't support rewinding the C stack with
delimited continuations.
* libguile/eval.c (eval): Adapt to scm_c_make_prompt change.
* libguile/vm-engine.c (vm_engine): Use vp->cookie to get a unique value
corresponding to this VM invocation.
* libguile/vm-i-system.c (prompt): Pass the cookie to scm_c_make_prompt.
(abort): Take an additional tail arg.
* libguile/vm.c (vm_abort): Parse out the abort tail arg. This is for
the @abort case, or the (apply abort ...) case.
(make_vm): Initialize the cookie to 0.
* libguile/vm.h (struct scm_vm): Add cookie.
* module/ice-9/boot-9.scm (abort): Define here as a trampoline to
@abort. Needed to make sure that a call to abort dispatches to a VM
opcode, so the cookie will be the same.
* module/language/tree-il.scm (<tree-il>): Add a "tail" field to
<abort>, for the (apply abort ...) case, or (@abort tag args). Should
be #<const ()> in the normal case. Add support throughout.
* module/language/tree-il/analyze.scm (analyze-lexicals): Add abort-tail
support here too.
* module/language/tree-il/compile-glil.scm (flatten): Compile the tail
argument appropriately.
* module/language/tree-il/primitives.scm (*primitive-expand-table*): Fix
@abort and abort cases to pass the tail arg to make-abort.
2010-02-22 21:53:24 +01:00
|
|
|
|
((<abort> tag args tail)
|
|
|
|
|
|
(apply lset-union eq? (step tag) (step tail) (map step args)))
|
2010-01-30 15:49:50 +01:00
|
|
|
|
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(else '())))
|
|
|
|
|
|
|
2009-08-07 15:35:53 +02:00
|
|
|
|
;; allocation: sym -> {lambda -> address}
|
|
|
|
|
|
;; lambda -> (nlocs labels . free-locs)
|
|
|
|
|
|
(define allocation (make-hash-table))
|
|
|
|
|
|
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(define (allocate! x proc n)
|
|
|
|
|
|
(define (recur y) (allocate! y proc n))
|
|
|
|
|
|
(record-case x
|
|
|
|
|
|
((<application> proc args)
|
|
|
|
|
|
(apply max (recur proc) (map recur args)))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-12-11 12:00:27 +01:00
|
|
|
|
((<conditional> test consequent alternate)
|
|
|
|
|
|
(max (recur test) (recur consequent) (recur alternate)))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<lexical-set> exp)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(recur exp))
|
|
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<module-set> exp)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(recur exp))
|
|
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<toplevel-set> exp)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(recur exp))
|
|
|
|
|
|
|
2009-09-21 00:35:19 +02:00
|
|
|
|
((<toplevel-define> exp)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(recur exp))
|
|
|
|
|
|
|
|
|
|
|
|
((<sequence> exps)
|
|
|
|
|
|
(apply max (map recur exps)))
|
|
|
|
|
|
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
((<lambda> body)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; allocate closure vars in order
|
|
|
|
|
|
(let lp ((c (hashq-ref free-vars x)) (n 0))
|
|
|
|
|
|
(if (pair? c)
|
|
|
|
|
|
(begin
|
|
|
|
|
|
(hashq-set! (hashq-ref allocation (car c))
|
|
|
|
|
|
x
|
|
|
|
|
|
`(#f ,(hashq-ref assigned (car c)) . ,n))
|
|
|
|
|
|
(lp (cdr c) (1+ n)))))
|
|
|
|
|
|
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
(let ((nlocs (allocate! body x 0))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(free-addresses
|
|
|
|
|
|
(map (lambda (v)
|
|
|
|
|
|
(hashq-ref (hashq-ref allocation v) proc))
|
2009-08-07 15:35:53 +02:00
|
|
|
|
(hashq-ref free-vars x)))
|
|
|
|
|
|
(labels (filter cdr
|
|
|
|
|
|
(map (lambda (sym)
|
|
|
|
|
|
(cons sym (hashq-ref labels sym)))
|
|
|
|
|
|
(hashq-ref bound-vars x)))))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
;; set procedure allocations
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
(hashq-set! allocation x (cons labels free-addresses)))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
n)
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-12-11 11:49:14 +01:00
|
|
|
|
((<lambda-case> opt kw inits vars body alternate)
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
(max
|
|
|
|
|
|
(let lp ((vars vars) (n n))
|
|
|
|
|
|
(if (null? vars)
|
2009-10-23 15:51:00 +02:00
|
|
|
|
(let ((nlocs (apply
|
|
|
|
|
|
max
|
|
|
|
|
|
(allocate! body proc n)
|
|
|
|
|
|
;; inits not logically at the end, but they
|
|
|
|
|
|
;; are the list...
|
2009-12-28 16:36:29 +01:00
|
|
|
|
(map (lambda (x) (allocate! x proc n)) inits))))
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
;; label and nlocs for the case
|
|
|
|
|
|
(hashq-set! allocation x (cons (gensym ":LCASE") nlocs))
|
|
|
|
|
|
nlocs)
|
|
|
|
|
|
(begin
|
|
|
|
|
|
(hashq-set! allocation (car vars)
|
|
|
|
|
|
(make-hashq
|
|
|
|
|
|
proc `(#t ,(hashq-ref assigned (car vars)) . ,n)))
|
|
|
|
|
|
(lp (cdr vars) (1+ n)))))
|
2009-12-11 11:49:14 +01:00
|
|
|
|
(if alternate (allocate! alternate proc n) n)))
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
|
2009-07-23 17:00:56 +02:00
|
|
|
|
((<let> vars vals body)
|
|
|
|
|
|
(let ((nmax (apply max (map recur vals))))
|
|
|
|
|
|
(cond
|
|
|
|
|
|
;; the `or' hack
|
|
|
|
|
|
((and (conditional? body)
|
|
|
|
|
|
(= (length vars) 1)
|
|
|
|
|
|
(let ((v (car vars)))
|
|
|
|
|
|
(and (not (hashq-ref assigned v))
|
|
|
|
|
|
(= (hashq-ref refcounts v 0) 2)
|
|
|
|
|
|
(lexical-ref? (conditional-test body))
|
|
|
|
|
|
(eq? (lexical-ref-gensym (conditional-test body)) v)
|
2009-12-11 12:00:27 +01:00
|
|
|
|
(lexical-ref? (conditional-consequent body))
|
|
|
|
|
|
(eq? (lexical-ref-gensym (conditional-consequent body)) v))))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(hashq-set! allocation (car vars)
|
|
|
|
|
|
(make-hashq proc `(#t #f . ,n)))
|
|
|
|
|
|
;; the 1+ for this var
|
2009-12-11 12:00:27 +01:00
|
|
|
|
(max nmax (1+ n) (allocate! (conditional-alternate body) proc n)))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(else
|
|
|
|
|
|
(let lp ((vars vars) (n n))
|
|
|
|
|
|
(if (null? vars)
|
|
|
|
|
|
(max nmax (allocate! body proc n))
|
|
|
|
|
|
(let ((v (car vars)))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
(hashq-set!
|
|
|
|
|
|
allocation v
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(make-hashq proc
|
|
|
|
|
|
`(#t ,(hashq-ref assigned v) . ,n)))
|
|
|
|
|
|
(lp (cdr vars) (1+ n)))))))))
|
|
|
|
|
|
|
|
|
|
|
|
((<letrec> vars vals body)
|
|
|
|
|
|
(let lp ((vars vars) (n n))
|
|
|
|
|
|
(if (null? vars)
|
|
|
|
|
|
(let ((nmax (apply max
|
|
|
|
|
|
(map (lambda (x)
|
|
|
|
|
|
(allocate! x proc n))
|
|
|
|
|
|
vals))))
|
|
|
|
|
|
(max nmax (allocate! body proc n)))
|
|
|
|
|
|
(let ((v (car vars)))
|
|
|
|
|
|
(hashq-set!
|
|
|
|
|
|
allocation v
|
|
|
|
|
|
(make-hashq proc
|
|
|
|
|
|
`(#t ,(hashq-ref assigned v) . ,n)))
|
|
|
|
|
|
(lp (cdr vars) (1+ n))))))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
add <fix> tree-il construct, and compile it
* libguile/vm-i-system.c (fix-closure): New instruction, for wiring
together fixpoint procedures.
* module/Makefile.am (TREE_IL_LANG_SOURCES): Add fix-letrec.scm.
* module/language/glil/compile-assembly.scm (glil->assembly): Reindent
the <glil-lexical> case, and handle 'fix for locally-bound vars.
* module/language/tree-il.scm (<fix>): Add the <fix> tree-il type and
accessors, for fixed-point bindings. This IL construct is taken from
the Waddell paper.
(parse-tree-il, unparse-tree-il, tree-il->scheme, tree-il-fold)
(pre-order!, post-order!): Update for <fix>.
* module/language/tree-il/analyze.scm (analyze-lexicals): Update for
<fix>. The difference here is that the bindings may not be assigned,
and are not marked as such. They are not boxed.
(report-unused-variables): Update for <fix>.
* module/language/tree-il/compile-glil.scm (flatten): Compile <fix> to
GLIL.
* module/language/tree-il/fix-letrec.scm: A stub implementation of
fixing letrec -- will flesh out in a separate commit.
* module/language/tree-il/inline.scm: Fix license, it was mistakenly
added with LGPL v2.1+.
* module/language/tree-il/optimize.scm (optimize!): Run the fix-letrec!
pass.
2009-08-05 17:51:40 +02:00
|
|
|
|
((<fix> vars vals body)
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(let lp ((in vars) (n n))
|
|
|
|
|
|
(if (null? in)
|
|
|
|
|
|
(let lp ((vars vars) (vals vals) (nmax n))
|
|
|
|
|
|
(cond
|
|
|
|
|
|
((null? vars)
|
|
|
|
|
|
(max nmax (allocate! body proc n)))
|
|
|
|
|
|
((hashq-ref labels (car vars))
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
;; allocate lambda body inline to proc
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(lp (cdr vars)
|
|
|
|
|
|
(cdr vals)
|
|
|
|
|
|
(record-case (car vals)
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
((<lambda> body)
|
|
|
|
|
|
(max nmax (allocate! body proc n))))))
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(else
|
|
|
|
|
|
;; allocate closure
|
|
|
|
|
|
(lp (cdr vars)
|
|
|
|
|
|
(cdr vals)
|
|
|
|
|
|
(max nmax (allocate! (car vals) proc n))))))
|
|
|
|
|
|
|
|
|
|
|
|
(let ((v (car in)))
|
|
|
|
|
|
(cond
|
|
|
|
|
|
((hashq-ref assigned v)
|
|
|
|
|
|
(error "fixpoint procedures may not be assigned" x))
|
|
|
|
|
|
((hashq-ref labels v)
|
|
|
|
|
|
;; no binding, it's a label
|
|
|
|
|
|
(lp (cdr in) n))
|
|
|
|
|
|
(else
|
|
|
|
|
|
;; allocate closure binding
|
|
|
|
|
|
(hashq-set! allocation v (make-hashq proc `(#t #f . ,n)))
|
|
|
|
|
|
(lp (cdr in) (1+ n))))))))
|
add <fix> tree-il construct, and compile it
* libguile/vm-i-system.c (fix-closure): New instruction, for wiring
together fixpoint procedures.
* module/Makefile.am (TREE_IL_LANG_SOURCES): Add fix-letrec.scm.
* module/language/glil/compile-assembly.scm (glil->assembly): Reindent
the <glil-lexical> case, and handle 'fix for locally-bound vars.
* module/language/tree-il.scm (<fix>): Add the <fix> tree-il type and
accessors, for fixed-point bindings. This IL construct is taken from
the Waddell paper.
(parse-tree-il, unparse-tree-il, tree-il->scheme, tree-il-fold)
(pre-order!, post-order!): Update for <fix>.
* module/language/tree-il/analyze.scm (analyze-lexicals): Update for
<fix>. The difference here is that the bindings may not be assigned,
and are not marked as such. They are not boxed.
(report-unused-variables): Update for <fix>.
* module/language/tree-il/compile-glil.scm (flatten): Compile <fix> to
GLIL.
* module/language/tree-il/fix-letrec.scm: A stub implementation of
fixing letrec -- will flesh out in a separate commit.
* module/language/tree-il/inline.scm: Fix license, it was mistakenly
added with LGPL v2.1+.
* module/language/tree-il/optimize.scm (optimize!): Run the fix-letrec!
pass.
2009-08-05 17:51:40 +02:00
|
|
|
|
|
tree-il support for case-lambda
* module/language/tree-il.scm (<lambda>, <lambda-case>): Split lambda
into the lambda itself, denoting the procedure, and lambda-case,
denoting a particular arity case. Lambda-case is fairly featureful,
and has not yet been fully tested.
(<let-values>): Use a <lambda-case> as the binding expression. Seems
to suit the purpose well.
Adapt parsers, unparsers, traversal operators, etc. Sometimes in this
first version we assume there are no optional args, rest args, or a
predicate.
* module/language/tree-il/analyze.scm (analyze-lexicals): Adapt for the
new case-lambda regime. Fairly well commented. It actually simplifies
things.
(report-unused-variables): Update for new tree-il.
* module/language/tree-il/compile-glil.scm: Adapt for the new tree-il.
There are some first stabs here at proper case-lambda compilation, but
they are untested as of yet.
* module/language/tree-il/inline.scm (inline!): Rework so we can
recurse on a single node; though these transformations are strictly
reductive, so they should complete in bounded time. Simplify
accordingly, and adapt to case-lambda. Oh, and we handle lambda->let
in not just the nullary case.
* module/ice-9/psyntax.scm (build-simple-lambda, build-case-lambda)
(build-lambda-case): New constructors. The idea is that after syntax
expansion, we shouldn't have to deal with improper lists any more.
Build-simple-lambda is a shortcut for the common case. The others are
not fully exercised yet. Adapt callers.
(syntax): Add some debugging in the lambda case. I don't fully
understand this, but in practice we don't seem to see rest args here.
(lambda): Inline chi-lambda-clause, and adapt for build-simple-lambda.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/tree-il.test: Update tests for new tree-il lambda
format, and to expect post-prelude labels for all glil programs.
2009-10-14 00:08:35 +02:00
|
|
|
|
((<let-values> exp body)
|
|
|
|
|
|
(max (recur exp) (recur body)))
|
2009-07-23 17:00:56 +02:00
|
|
|
|
|
2010-02-18 17:23:49 +01:00
|
|
|
|
((<dynwind> body winder unwinder)
|
2010-01-30 15:49:50 +01:00
|
|
|
|
(max (recur body) (recur winder) (recur unwinder)))
|
|
|
|
|
|
|
add dynlet to tree-il
* module/language/tree-il.scm (<dynlet>, dynlet?, make-dynlet)
(dynlet-src, dynlet-fluids, dynlet-vals, dynlet-body): New tree-il
construct, mapping to `with-fluids'.
(parse-tree-il, unparse-tree-il, tree-il->scheme, tree-il-fold):
(make-tree-il-folder, post-order!, pre-order!): Wire it up.
* module/language/tree-il/analyze.scm (analyze-lexicals): Add dynlet
support.
2010-02-18 18:05:24 +01:00
|
|
|
|
((<dynlet> fluids vals body)
|
|
|
|
|
|
(apply max (recur body) (map recur (append fluids vals))))
|
|
|
|
|
|
|
2010-02-19 11:42:00 +01:00
|
|
|
|
((<dynref> fluid)
|
|
|
|
|
|
(recur fluid))
|
|
|
|
|
|
|
|
|
|
|
|
((<dynset> fluid exp)
|
|
|
|
|
|
(max (recur fluid) (recur exp)))
|
|
|
|
|
|
|
2010-02-18 23:56:12 +01:00
|
|
|
|
((<prompt> tag body handler)
|
2010-01-30 15:49:50 +01:00
|
|
|
|
(let ((cont-var (and (lambda-case? handler)
|
|
|
|
|
|
(pair? (lambda-case-vars handler))
|
|
|
|
|
|
(car (lambda-case-vars handler)))))
|
|
|
|
|
|
(hashq-set! allocation x
|
|
|
|
|
|
(and cont-var (zero? (hashq-ref refcounts cont-var 0))))
|
2010-02-18 23:56:12 +01:00
|
|
|
|
(max (recur tag) (recur body) (recur handler))))
|
2010-01-30 15:49:50 +01:00
|
|
|
|
|
abort always dispatches to VM bytecode, to detect same-invocation aborts
* libguile/control.h:
* libguile/control.c (scm_c_make_prompt): Take an extra arg, a cookie.
Continuations will be rewindable only if the abort has the same cookie
as the prompt.
(scm_at_abort): Redefine from scm_abort, and instead of taking rest
args, take the abort values as a list directly. Also, don't allow
rewinding, because we won't support rewinding the C stack with
delimited continuations.
* libguile/eval.c (eval): Adapt to scm_c_make_prompt change.
* libguile/vm-engine.c (vm_engine): Use vp->cookie to get a unique value
corresponding to this VM invocation.
* libguile/vm-i-system.c (prompt): Pass the cookie to scm_c_make_prompt.
(abort): Take an additional tail arg.
* libguile/vm.c (vm_abort): Parse out the abort tail arg. This is for
the @abort case, or the (apply abort ...) case.
(make_vm): Initialize the cookie to 0.
* libguile/vm.h (struct scm_vm): Add cookie.
* module/ice-9/boot-9.scm (abort): Define here as a trampoline to
@abort. Needed to make sure that a call to abort dispatches to a VM
opcode, so the cookie will be the same.
* module/language/tree-il.scm (<tree-il>): Add a "tail" field to
<abort>, for the (apply abort ...) case, or (@abort tag args). Should
be #<const ()> in the normal case. Add support throughout.
* module/language/tree-il/analyze.scm (analyze-lexicals): Add abort-tail
support here too.
* module/language/tree-il/compile-glil.scm (flatten): Compile the tail
argument appropriately.
* module/language/tree-il/primitives.scm (*primitive-expand-table*): Fix
@abort and abort cases to pass the tail arg to make-abort.
2010-02-22 21:53:24 +01:00
|
|
|
|
((<abort> tag args tail)
|
|
|
|
|
|
(apply max (recur tag) (recur tail) (map recur args)))
|
2010-01-30 15:49:50 +01:00
|
|
|
|
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(else n)))
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
2009-08-07 19:06:15 +02:00
|
|
|
|
(analyze! x #f '() #t #f)
|
2009-07-23 17:00:56 +02:00
|
|
|
|
(allocate! x #f 0)
|
2009-05-15 23:44:14 +02:00
|
|
|
|
|
|
|
|
|
|
allocation)
|
2009-07-31 00:42:58 +02:00
|
|
|
|
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
|
|
|
|
|
;;;
|
|
|
|
|
|
;;; Tree analyses for warnings.
|
|
|
|
|
|
;;;
|
|
|
|
|
|
|
|
|
|
|
|
(define-record-type <tree-analysis>
|
|
|
|
|
|
(make-tree-analysis leaf down up post init)
|
|
|
|
|
|
tree-analysis?
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(leaf tree-analysis-leaf) ;; (lambda (x result env locs) ...)
|
|
|
|
|
|
(down tree-analysis-down) ;; (lambda (x result env locs) ...)
|
|
|
|
|
|
(up tree-analysis-up) ;; (lambda (x result env locs) ...)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(post tree-analysis-post) ;; (lambda (result env) ...)
|
|
|
|
|
|
(init tree-analysis-init)) ;; arbitrary value
|
|
|
|
|
|
|
|
|
|
|
|
(define (analyze-tree analyses tree env)
|
|
|
|
|
|
"Run all tree analyses listed in ANALYSES on TREE for ENV, using
|
2010-01-08 12:02:00 +01:00
|
|
|
|
`tree-il-fold'. Return TREE. The leaf/down/up procedures of each analysis are
|
|
|
|
|
|
passed a ``location stack', which is the stack of `tree-il-src' values for each
|
|
|
|
|
|
parent tree (a list); it can be used to approximate source location when
|
|
|
|
|
|
accurate information is missing from a given `tree-il' element."
|
|
|
|
|
|
|
|
|
|
|
|
(define (traverse proc update-locs)
|
|
|
|
|
|
;; Return a tree traversing procedure that returns a list of analysis
|
|
|
|
|
|
;; results prepended by the location stack.
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(lambda (x results)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(let ((locs (update-locs x (car results))))
|
|
|
|
|
|
(cons locs ;; the location stack
|
|
|
|
|
|
(map (lambda (analysis result)
|
|
|
|
|
|
((proc analysis) x result env locs))
|
|
|
|
|
|
analyses
|
|
|
|
|
|
(cdr results))))))
|
|
|
|
|
|
|
|
|
|
|
|
;; Keeping/extending/shrinking the location stack.
|
|
|
|
|
|
(define (keep-locs x locs) locs)
|
|
|
|
|
|
(define (extend-locs x locs) (cons (tree-il-src x) locs))
|
|
|
|
|
|
(define (shrink-locs x locs) (cdr locs))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
|
|
|
|
|
(let ((results
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(tree-il-fold (traverse tree-analysis-leaf keep-locs)
|
|
|
|
|
|
(traverse tree-analysis-down extend-locs)
|
|
|
|
|
|
(traverse tree-analysis-up shrink-locs)
|
|
|
|
|
|
(cons '() ;; empty location stack
|
|
|
|
|
|
(map tree-analysis-init analyses))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
tree)))
|
|
|
|
|
|
|
|
|
|
|
|
(for-each (lambda (analysis result)
|
|
|
|
|
|
((tree-analysis-post analysis) result env))
|
|
|
|
|
|
analyses
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(cdr results)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
|
|
|
|
|
tree)
|
|
|
|
|
|
|
2009-07-31 00:42:58 +02:00
|
|
|
|
|
|
|
|
|
|
;;;
|
|
|
|
|
|
;;; Unused variable analysis.
|
|
|
|
|
|
;;;
|
|
|
|
|
|
|
|
|
|
|
|
;; <binding-info> records are used during tree traversals in
|
2010-01-08 12:02:00 +01:00
|
|
|
|
;; `unused-variable-analysis'. They contain a list of the local vars
|
|
|
|
|
|
;; currently in scope, and a list of locals vars that have been referenced.
|
2009-07-31 00:42:58 +02:00
|
|
|
|
(define-record-type <binding-info>
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info vars refs)
|
2009-07-31 00:42:58 +02:00
|
|
|
|
binding-info?
|
|
|
|
|
|
(vars binding-info-vars) ;; ((GENSYM NAME LOCATION) ...)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(refs binding-info-refs)) ;; (GENSYM ...)
|
2009-07-31 00:42:58 +02:00
|
|
|
|
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(define unused-variable-analysis
|
2009-11-07 18:32:26 +01:00
|
|
|
|
;; Report unused variables in the given tree.
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(make-tree-analysis
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
;; X is a leaf: extend INFO's refs accordingly.
|
|
|
|
|
|
(let ((refs (binding-info-refs info))
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(vars (binding-info-vars info)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(record-case x
|
|
|
|
|
|
((<lexical-ref> gensym)
|
2010-02-02 23:58:03 +01:00
|
|
|
|
(make-binding-info vars (vhash-consq gensym #t refs)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(else info))))
|
|
|
|
|
|
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
;; Going down into X: extend INFO's variable list
|
|
|
|
|
|
;; accordingly.
|
|
|
|
|
|
(let ((refs (binding-info-refs info))
|
|
|
|
|
|
(vars (binding-info-vars info))
|
|
|
|
|
|
(src (tree-il-src x)))
|
|
|
|
|
|
(define (extend inner-vars inner-names)
|
2010-02-02 23:58:03 +01:00
|
|
|
|
(fold (lambda (var name vars)
|
|
|
|
|
|
(vhash-consq var (list name src) vars))
|
|
|
|
|
|
vars
|
|
|
|
|
|
inner-vars
|
|
|
|
|
|
inner-names))
|
|
|
|
|
|
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(record-case x
|
|
|
|
|
|
((<lexical-set> gensym)
|
2010-02-02 23:58:03 +01:00
|
|
|
|
(make-binding-info vars (vhash-consq gensym #t refs)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
((<lambda-case> req opt inits rest kw vars)
|
|
|
|
|
|
(let ((names `(,@req
|
2009-11-08 01:08:54 +01:00
|
|
|
|
,@(or opt '())
|
2009-11-06 10:42:45 +01:00
|
|
|
|
,@(if rest (list rest) '())
|
|
|
|
|
|
,@(if kw (map cadr (cdr kw)) '()))))
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info (extend vars names) refs)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
((<let> vars names)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info (extend vars names) refs))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
((<letrec> vars names)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info (extend vars names) refs))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
((<fix> vars names)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info (extend vars names) refs))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(else info))))
|
|
|
|
|
|
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
;; Leaving X's scope: shrink INFO's variable list
|
|
|
|
|
|
;; accordingly and reported unused nested variables.
|
|
|
|
|
|
(let ((refs (binding-info-refs info))
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(vars (binding-info-vars info)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(define (shrink inner-vars refs)
|
2010-02-02 23:58:03 +01:00
|
|
|
|
(vlist-for-each
|
|
|
|
|
|
(lambda (var)
|
|
|
|
|
|
(let ((gensym (car var)))
|
|
|
|
|
|
;; Don't report lambda parameters as unused.
|
|
|
|
|
|
(if (and (memq gensym inner-vars)
|
|
|
|
|
|
(not (vhash-assq gensym refs))
|
|
|
|
|
|
(not (lambda-case? x)))
|
|
|
|
|
|
(let ((name (cadr var))
|
|
|
|
|
|
;; We can get approximate source location by going up
|
|
|
|
|
|
;; the LOCS location stack.
|
|
|
|
|
|
(loc (or (caddr var)
|
|
|
|
|
|
(find pair? locs))))
|
|
|
|
|
|
(warning 'unused-variable loc name)))))
|
|
|
|
|
|
vars)
|
|
|
|
|
|
(vlist-drop vars (length inner-vars)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
|
|
|
|
|
;; For simplicity, we leave REFS untouched, i.e., with
|
|
|
|
|
|
;; names of variables that are now going out of scope.
|
|
|
|
|
|
;; It doesn't hurt as these are unique names, it just
|
|
|
|
|
|
;; makes REFS unnecessarily fat.
|
|
|
|
|
|
(record-case x
|
|
|
|
|
|
((<lambda-case> vars)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info (shrink vars refs) refs))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
((<let> vars)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info (shrink vars refs) refs))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
((<letrec> vars)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info (shrink vars refs) refs))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
((<fix> vars)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-binding-info (shrink vars refs) refs))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(else info))))
|
|
|
|
|
|
|
|
|
|
|
|
(lambda (result env) #t)
|
2010-02-02 23:58:03 +01:00
|
|
|
|
(make-binding-info vlist-null vlist-null)))
|
2009-10-06 23:39:56 +02:00
|
|
|
|
|
2010-01-11 01:19:16 +01:00
|
|
|
|
|
|
|
|
|
|
;;;
|
|
|
|
|
|
;;; Unused top-level variable analysis.
|
|
|
|
|
|
;;;
|
|
|
|
|
|
|
2010-01-11 18:28:19 +01:00
|
|
|
|
;; <reference-graph> record top-level definitions that are made, references to
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;; top-level definitions and their context (the top-level definition in which
|
|
|
|
|
|
;; the reference appears), as well as the current context (the top-level
|
|
|
|
|
|
;; definition we're currently in). The second part (`refs' below) is
|
2010-01-11 18:28:19 +01:00
|
|
|
|
;; effectively a graph from which we can determine unused top-level definitions.
|
|
|
|
|
|
(define-record-type <reference-graph>
|
|
|
|
|
|
(make-reference-graph refs defs toplevel-context)
|
|
|
|
|
|
reference-graph?
|
|
|
|
|
|
(defs reference-graph-defs) ;; ((NAME . LOC) ...)
|
|
|
|
|
|
(refs reference-graph-refs) ;; ((REF-CONTEXT REF ...) ...)
|
|
|
|
|
|
(toplevel-context reference-graph-toplevel-context)) ;; NAME | #f
|
|
|
|
|
|
|
2010-02-02 23:59:03 +01:00
|
|
|
|
(define (graph-reachable-nodes root refs reachable)
|
|
|
|
|
|
;; Add to REACHABLE the nodes reachable from ROOT in graph REFS. REFS is a
|
|
|
|
|
|
;; vhash mapping nodes to the list of their children: for instance,
|
|
|
|
|
|
;; ((A -> (B C)) (B -> (A)) (C -> ())) corresponds to
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;;
|
|
|
|
|
|
;; ,-------.
|
|
|
|
|
|
;; v |
|
|
|
|
|
|
;; A ----> B
|
|
|
|
|
|
;; |
|
|
|
|
|
|
;; v
|
|
|
|
|
|
;; C
|
2010-02-02 23:59:03 +01:00
|
|
|
|
;;
|
|
|
|
|
|
;; REACHABLE is a vhash of nodes known to be otherwise reachable.
|
2010-01-11 01:19:16 +01:00
|
|
|
|
|
|
|
|
|
|
(let loop ((root root)
|
2010-02-02 23:59:03 +01:00
|
|
|
|
(path vlist-null)
|
|
|
|
|
|
(result reachable))
|
|
|
|
|
|
(if (or (vhash-assq root path)
|
|
|
|
|
|
(vhash-assq root result))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
result
|
2010-02-02 23:59:03 +01:00
|
|
|
|
(let* ((children (or (and=> (vhash-assq root refs) cdr) '()))
|
|
|
|
|
|
(path (vhash-consq root #t path))
|
|
|
|
|
|
(result (fold (lambda (kid result)
|
|
|
|
|
|
(loop kid path result))
|
|
|
|
|
|
result
|
|
|
|
|
|
children)))
|
|
|
|
|
|
(fold (lambda (kid result)
|
|
|
|
|
|
(vhash-consq kid #t result))
|
|
|
|
|
|
result
|
|
|
|
|
|
children)))))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(define (graph-reachable-nodes* roots refs)
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;; Return the list of nodes in REFS reachable from the nodes listed in ROOTS.
|
2010-02-02 23:59:03 +01:00
|
|
|
|
(vlist-fold (lambda (root+true result)
|
|
|
|
|
|
(let* ((root (car root+true))
|
|
|
|
|
|
(reachable (graph-reachable-nodes root refs result)))
|
|
|
|
|
|
(vhash-consq root #t reachable)))
|
|
|
|
|
|
vlist-null
|
|
|
|
|
|
roots))
|
|
|
|
|
|
|
|
|
|
|
|
(define (partition* pred vhash)
|
|
|
|
|
|
;; Partition VHASH according to PRED. Return the two resulting vhashes.
|
|
|
|
|
|
(let ((result
|
|
|
|
|
|
(vlist-fold (lambda (k+v result)
|
|
|
|
|
|
(let ((k (car k+v))
|
|
|
|
|
|
(v (cdr k+v))
|
|
|
|
|
|
(r1 (car result))
|
|
|
|
|
|
(r2 (cdr result)))
|
|
|
|
|
|
(if (pred k)
|
|
|
|
|
|
(cons (vhash-consq k v r1) r2)
|
|
|
|
|
|
(cons r1 (vhash-consq k v r2)))))
|
|
|
|
|
|
(cons vlist-null vlist-null)
|
|
|
|
|
|
vhash)))
|
|
|
|
|
|
(values (car result) (cdr result))))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
|
|
|
|
|
|
(define unused-toplevel-analysis
|
|
|
|
|
|
;; Report unused top-level definitions that are not exported.
|
|
|
|
|
|
(let ((add-ref-from-context
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(lambda (graph name)
|
|
|
|
|
|
;; Add an edge CTX -> NAME in GRAPH.
|
|
|
|
|
|
(let* ((refs (reference-graph-refs graph))
|
|
|
|
|
|
(defs (reference-graph-defs graph))
|
|
|
|
|
|
(ctx (reference-graph-toplevel-context graph))
|
2010-02-02 23:59:03 +01:00
|
|
|
|
(ctx-refs (or (and=> (vhash-assq ctx refs) cdr) '())))
|
|
|
|
|
|
(make-reference-graph (vhash-consq ctx (cons name ctx-refs) refs)
|
2010-01-11 18:28:19 +01:00
|
|
|
|
defs ctx)))))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
(define (macro-variable? name env)
|
|
|
|
|
|
(and (module? env)
|
|
|
|
|
|
(let ((var (module-variable env name)))
|
|
|
|
|
|
(and var (variable-bound? var)
|
|
|
|
|
|
(macro? (variable-ref var))))))
|
|
|
|
|
|
|
|
|
|
|
|
(make-tree-analysis
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(lambda (x graph env locs)
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;; X is a leaf.
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(let ((ctx (reference-graph-toplevel-context graph)))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
(record-case x
|
|
|
|
|
|
((<toplevel-ref> name src)
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(add-ref-from-context graph name))
|
|
|
|
|
|
(else graph))))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(lambda (x graph env locs)
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;; Going down into X.
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(let ((ctx (reference-graph-toplevel-context graph))
|
|
|
|
|
|
(refs (reference-graph-refs graph))
|
|
|
|
|
|
(defs (reference-graph-defs graph)))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
(record-case x
|
|
|
|
|
|
((<toplevel-define> name src)
|
|
|
|
|
|
(let ((refs refs)
|
2010-02-02 23:59:03 +01:00
|
|
|
|
(defs (vhash-consq name (or src (find pair? locs))
|
|
|
|
|
|
defs)))
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(make-reference-graph refs defs name)))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
((<toplevel-set> name src)
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(add-ref-from-context graph name))
|
|
|
|
|
|
(else graph))))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(lambda (x graph env locs)
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;; Leaving X's scope.
|
|
|
|
|
|
(record-case x
|
|
|
|
|
|
((<toplevel-define>)
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(let ((refs (reference-graph-refs graph))
|
|
|
|
|
|
(defs (reference-graph-defs graph)))
|
|
|
|
|
|
(make-reference-graph refs defs #f)))
|
|
|
|
|
|
(else graph)))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(lambda (graph env)
|
|
|
|
|
|
;; Process the resulting reference graph: determine all private definitions
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;; not reachable from any public definition. Macros
|
|
|
|
|
|
;; (syntax-transformers), which are globally bound, never considered
|
|
|
|
|
|
;; unused since we can't tell whether a macro is actually used; in
|
2010-01-11 18:28:19 +01:00
|
|
|
|
;; addition, macros are considered roots of the graph since they may use
|
2010-01-11 01:19:16 +01:00
|
|
|
|
;; private bindings. FIXME: The `make-syntax-transformer' calls don't
|
|
|
|
|
|
;; contain any literal `toplevel-ref' of the global bindings they use so
|
|
|
|
|
|
;; this strategy fails.
|
|
|
|
|
|
(define (exported? name)
|
|
|
|
|
|
(if (module? env)
|
|
|
|
|
|
(module-variable (module-public-interface env) name)
|
|
|
|
|
|
#t))
|
|
|
|
|
|
|
|
|
|
|
|
(let-values (((public-defs private-defs)
|
2010-02-02 23:59:03 +01:00
|
|
|
|
(partition* (lambda (name)
|
|
|
|
|
|
(or (exported? name)
|
|
|
|
|
|
(macro-variable? name env)))
|
|
|
|
|
|
(reference-graph-defs graph))))
|
|
|
|
|
|
(let* ((roots (vhash-consq #f #t public-defs))
|
2010-01-11 18:28:19 +01:00
|
|
|
|
(refs (reference-graph-refs graph))
|
|
|
|
|
|
(reachable (graph-reachable-nodes* roots refs))
|
2010-02-02 23:59:03 +01:00
|
|
|
|
(unused (vlist-filter (lambda (name+src)
|
|
|
|
|
|
(not (vhash-assq (car name+src)
|
|
|
|
|
|
reachable)))
|
|
|
|
|
|
private-defs)))
|
|
|
|
|
|
(vlist-for-each (lambda (name+loc)
|
|
|
|
|
|
(let ((name (car name+loc))
|
|
|
|
|
|
(loc (cdr name+loc)))
|
|
|
|
|
|
(warning 'unused-toplevel loc name)))
|
|
|
|
|
|
unused))))
|
|
|
|
|
|
|
|
|
|
|
|
(make-reference-graph vlist-null vlist-null #f))))
|
2010-01-11 01:19:16 +01:00
|
|
|
|
|
2009-10-06 23:39:56 +02:00
|
|
|
|
|
|
|
|
|
|
;;;
|
|
|
|
|
|
;;; Unbound variable analysis.
|
|
|
|
|
|
;;;
|
|
|
|
|
|
|
|
|
|
|
|
;; <toplevel-info> records are used during tree traversal in search of
|
|
|
|
|
|
;; possibly unbound variable. They contain a list of references to
|
2010-01-08 12:02:00 +01:00
|
|
|
|
;; potentially unbound top-level variables, and a list of the top-level
|
|
|
|
|
|
;; defines that have been encountered.
|
2009-10-06 23:39:56 +02:00
|
|
|
|
(define-record-type <toplevel-info>
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-toplevel-info refs defs)
|
2009-10-06 23:39:56 +02:00
|
|
|
|
toplevel-info?
|
|
|
|
|
|
(refs toplevel-info-refs) ;; ((VARIABLE-NAME . LOCATION) ...)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(defs toplevel-info-defs)) ;; (VARIABLE-NAME ...)
|
2009-10-06 23:39:56 +02:00
|
|
|
|
|
2009-10-22 22:33:53 +02:00
|
|
|
|
(define (goops-toplevel-definition proc args env)
|
2009-10-22 00:37:36 +02:00
|
|
|
|
;; If application of PROC to ARGS is a GOOPS top-level definition, return
|
|
|
|
|
|
;; the name of the variable being defined; otherwise return #f. This
|
|
|
|
|
|
;; assumes knowledge of the current implementation of `define-class' et al.
|
2009-10-22 22:33:53 +02:00
|
|
|
|
(define (toplevel-define-arg args)
|
|
|
|
|
|
(and (pair? args) (pair? (cdr args)) (null? (cddr args))
|
|
|
|
|
|
(record-case (car args)
|
|
|
|
|
|
((<const> exp)
|
|
|
|
|
|
(and (symbol? exp) exp))
|
|
|
|
|
|
(else #f))))
|
|
|
|
|
|
|
2009-10-22 00:37:36 +02:00
|
|
|
|
(record-case proc
|
|
|
|
|
|
((<module-ref> mod public? name)
|
|
|
|
|
|
(and (equal? mod '(oop goops))
|
|
|
|
|
|
(not public?)
|
|
|
|
|
|
(eq? name 'toplevel-define!)
|
2009-10-22 22:33:53 +02:00
|
|
|
|
(toplevel-define-arg args)))
|
|
|
|
|
|
((<toplevel-ref> name)
|
|
|
|
|
|
;; This may be the result of expanding one of the GOOPS macros within
|
|
|
|
|
|
;; `oop/goops.scm'.
|
|
|
|
|
|
(and (eq? name 'toplevel-define!)
|
|
|
|
|
|
(eq? env (resolve-module '(oop goops)))
|
|
|
|
|
|
(toplevel-define-arg args)))
|
2009-10-22 00:37:36 +02:00
|
|
|
|
(else #f)))
|
|
|
|
|
|
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(define unbound-variable-analysis
|
2009-11-07 18:32:26 +01:00
|
|
|
|
;; Report possibly unbound variables in the given tree.
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(make-tree-analysis
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
;; X is a leaf: extend INFO's refs accordingly.
|
|
|
|
|
|
(let ((refs (toplevel-info-refs info))
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(defs (toplevel-info-defs info)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(define (bound? name)
|
|
|
|
|
|
(or (and (module? env)
|
|
|
|
|
|
(module-variable env name))
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(vhash-assq name defs)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
|
|
|
|
|
(record-case x
|
|
|
|
|
|
((<toplevel-ref> name src)
|
|
|
|
|
|
(if (bound? name)
|
|
|
|
|
|
info
|
|
|
|
|
|
(let ((src (or src (find pair? locs))))
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(make-toplevel-info (vhash-consq name src refs)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
defs))))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(else info))))
|
|
|
|
|
|
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
;; Going down into X.
|
|
|
|
|
|
(let* ((refs (toplevel-info-refs info))
|
|
|
|
|
|
(defs (toplevel-info-defs info))
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(src (tree-il-src x)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(define (bound? name)
|
|
|
|
|
|
(or (and (module? env)
|
|
|
|
|
|
(module-variable env name))
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(vhash-assq name defs)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
|
|
|
|
|
(record-case x
|
|
|
|
|
|
((<toplevel-set> name src)
|
|
|
|
|
|
(if (bound? name)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-toplevel-info refs defs)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(let ((src (find pair? locs)))
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(make-toplevel-info (vhash-consq name src refs)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
defs))))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
((<toplevel-define> name)
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(make-toplevel-info (vhash-delete name refs eq?)
|
|
|
|
|
|
(vhash-consq name #t defs)))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
|
|
|
|
|
((<application> proc args)
|
|
|
|
|
|
;; Check for a dynamic top-level definition, as is
|
|
|
|
|
|
;; done by code expanded from GOOPS macros.
|
|
|
|
|
|
(let ((name (goops-toplevel-definition proc args
|
|
|
|
|
|
env)))
|
|
|
|
|
|
(if (symbol? name)
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(make-toplevel-info (vhash-delete name refs
|
2009-11-06 10:42:45 +01:00
|
|
|
|
eq?)
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(vhash-consq name #t defs))
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-toplevel-info refs defs))))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
(else
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(make-toplevel-info refs defs)))))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
;; Leaving X's scope.
|
2010-01-11 01:19:16 +01:00
|
|
|
|
info)
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
|
|
|
|
|
(lambda (toplevel env)
|
|
|
|
|
|
;; Post-process the result.
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(vlist-for-each (lambda (name+loc)
|
|
|
|
|
|
(let ((name (car name+loc))
|
|
|
|
|
|
(loc (cdr name+loc)))
|
|
|
|
|
|
(warning 'unbound-variable loc name)))
|
|
|
|
|
|
(vlist-reverse (toplevel-info-refs toplevel))))
|
2009-11-06 10:42:45 +01:00
|
|
|
|
|
2010-02-02 23:59:34 +01:00
|
|
|
|
(make-toplevel-info vlist-null vlist-null)))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
;;;
|
|
|
|
|
|
;;; Arity analysis.
|
|
|
|
|
|
;;;
|
|
|
|
|
|
|
2009-11-08 01:02:08 +01:00
|
|
|
|
;; <arity-info> records contain information about lexical definitions of
|
2009-11-07 18:32:26 +01:00
|
|
|
|
;; procedures currently in scope, top-level procedure definitions that have
|
|
|
|
|
|
;; been encountered, and calls to top-level procedures that have been
|
|
|
|
|
|
;; encountered.
|
|
|
|
|
|
(define-record-type <arity-info>
|
|
|
|
|
|
(make-arity-info toplevel-calls lexical-lambdas toplevel-lambdas)
|
|
|
|
|
|
arity-info?
|
|
|
|
|
|
(toplevel-calls toplevel-procedure-calls) ;; ((NAME . APPLICATION) ...)
|
|
|
|
|
|
(lexical-lambdas lexical-lambdas) ;; ((GENSYM . DEFINITION) ...)
|
|
|
|
|
|
(toplevel-lambdas toplevel-lambdas)) ;; ((NAME . DEFINITION) ...)
|
|
|
|
|
|
|
|
|
|
|
|
(define (validate-arity proc application lexical?)
|
|
|
|
|
|
;; Validate the argument count of APPLICATION, a tree-il application of
|
|
|
|
|
|
;; PROC, emitting a warning in case of argument count mismatch.
|
|
|
|
|
|
|
2009-11-08 01:02:08 +01:00
|
|
|
|
(define (filter-keyword-args keywords allow-other-keys? args)
|
|
|
|
|
|
;; Filter keyword arguments from ARGS and return the resulting list.
|
|
|
|
|
|
;; KEYWORDS is the list of allowed keywords, and ALLOW-OTHER-KEYS?
|
|
|
|
|
|
;; specified whethere keywords not listed in KEYWORDS are allowed.
|
|
|
|
|
|
(let loop ((args args)
|
|
|
|
|
|
(result '()))
|
|
|
|
|
|
(if (null? args)
|
|
|
|
|
|
(reverse result)
|
|
|
|
|
|
(let ((arg (car args)))
|
|
|
|
|
|
(if (and (const? arg)
|
|
|
|
|
|
(or (memq (const-exp arg) keywords)
|
|
|
|
|
|
(and allow-other-keys?
|
|
|
|
|
|
(keyword? (const-exp arg)))))
|
|
|
|
|
|
(loop (if (pair? (cdr args))
|
|
|
|
|
|
(cddr args)
|
|
|
|
|
|
'())
|
|
|
|
|
|
result)
|
|
|
|
|
|
(loop (cdr args)
|
|
|
|
|
|
(cons arg result)))))))
|
|
|
|
|
|
|
2009-11-08 17:53:14 +01:00
|
|
|
|
(define (arities proc)
|
|
|
|
|
|
;; Return the arities of PROC, which can be either a tree-il or a
|
2009-11-07 18:32:26 +01:00
|
|
|
|
;; procedure.
|
|
|
|
|
|
(define (len x)
|
|
|
|
|
|
(or (and (or (null? x) (pair? x))
|
|
|
|
|
|
(length x))
|
|
|
|
|
|
0))
|
2009-11-08 01:02:08 +01:00
|
|
|
|
(cond ((program? proc)
|
2010-04-17 15:17:24 +02:00
|
|
|
|
(values (procedure-name proc)
|
2009-11-08 17:53:14 +01:00
|
|
|
|
(map (lambda (a)
|
|
|
|
|
|
(list (arity:nreq a) (arity:nopt a) (arity:rest? a)
|
|
|
|
|
|
(map car (arity:kw a))
|
|
|
|
|
|
(arity:allow-other-keys? a)))
|
|
|
|
|
|
(program-arities proc))))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
((procedure? proc)
|
2010-04-17 16:28:52 +02:00
|
|
|
|
(let ((arity (procedure-minimum-arity proc)))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
(values (procedure-name proc)
|
2009-11-08 17:53:14 +01:00
|
|
|
|
(list (list (car arity) (cadr arity) (caddr arity)
|
|
|
|
|
|
#f #f)))))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
(else
|
2009-11-08 17:53:14 +01:00
|
|
|
|
(let loop ((name #f)
|
|
|
|
|
|
(proc proc)
|
|
|
|
|
|
(arities '()))
|
|
|
|
|
|
(if (not proc)
|
|
|
|
|
|
(values name (reverse arities))
|
|
|
|
|
|
(record-case proc
|
2009-12-11 11:49:14 +01:00
|
|
|
|
((<lambda-case> req opt rest kw alternate)
|
|
|
|
|
|
(loop name alternate
|
2009-11-08 17:53:14 +01:00
|
|
|
|
(cons (list (len req) (len opt) rest
|
|
|
|
|
|
(and (pair? kw) (map car (cdr kw)))
|
|
|
|
|
|
(and (pair? kw) (car kw)))
|
|
|
|
|
|
arities)))
|
|
|
|
|
|
((<lambda> meta body)
|
|
|
|
|
|
(loop (assoc-ref meta 'name) body arities))
|
|
|
|
|
|
(else
|
|
|
|
|
|
(values #f #f))))))))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
|
|
|
|
|
|
(let ((args (application-args application))
|
|
|
|
|
|
(src (tree-il-src application)))
|
2009-11-08 17:53:14 +01:00
|
|
|
|
(call-with-values (lambda () (arities proc))
|
|
|
|
|
|
(lambda (name arities)
|
|
|
|
|
|
(define matches?
|
|
|
|
|
|
(find (lambda (arity)
|
|
|
|
|
|
(pmatch arity
|
|
|
|
|
|
((,req ,opt ,rest? ,kw ,aok?)
|
|
|
|
|
|
(let ((args (if (pair? kw)
|
|
|
|
|
|
(filter-keyword-args kw aok? args)
|
|
|
|
|
|
args)))
|
|
|
|
|
|
(if (and req opt)
|
|
|
|
|
|
(let ((count (length args)))
|
|
|
|
|
|
(and (>= count req)
|
|
|
|
|
|
(or rest?
|
|
|
|
|
|
(<= count (+ req opt)))))
|
|
|
|
|
|
#t)))
|
|
|
|
|
|
(else #t)))
|
|
|
|
|
|
arities))
|
|
|
|
|
|
|
|
|
|
|
|
(if (not matches?)
|
|
|
|
|
|
(warning 'arity-mismatch src
|
|
|
|
|
|
(or name (with-output-to-string (lambda () (write proc))))
|
|
|
|
|
|
lexical?)))))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
#t)
|
|
|
|
|
|
|
|
|
|
|
|
(define arity-analysis
|
|
|
|
|
|
;; Report arity mismatches in the given tree.
|
|
|
|
|
|
(make-tree-analysis
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-07 18:32:26 +01:00
|
|
|
|
;; X is a leaf.
|
|
|
|
|
|
info)
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-07 18:32:26 +01:00
|
|
|
|
;; Down into X.
|
|
|
|
|
|
(define (extend lexical-name val info)
|
|
|
|
|
|
;; If VAL is a lambda, add NAME to the lexical-lambdas of INFO.
|
|
|
|
|
|
(let ((toplevel-calls (toplevel-procedure-calls info))
|
|
|
|
|
|
(lexical-lambdas (lexical-lambdas info))
|
|
|
|
|
|
(toplevel-lambdas (toplevel-lambdas info)))
|
|
|
|
|
|
(record-case val
|
|
|
|
|
|
((<lambda> body)
|
|
|
|
|
|
(make-arity-info toplevel-calls
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(vhash-consq lexical-name val
|
|
|
|
|
|
lexical-lambdas)
|
2009-11-07 18:32:26 +01:00
|
|
|
|
toplevel-lambdas))
|
|
|
|
|
|
((<lexical-ref> gensym)
|
|
|
|
|
|
;; lexical alias
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(let ((val* (vhash-assq gensym lexical-lambdas)))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
(if (pair? val*)
|
|
|
|
|
|
(extend lexical-name (cdr val*) info)
|
|
|
|
|
|
info)))
|
|
|
|
|
|
((<toplevel-ref> name)
|
|
|
|
|
|
;; top-level alias
|
|
|
|
|
|
(make-arity-info toplevel-calls
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(vhash-consq lexical-name val
|
|
|
|
|
|
lexical-lambdas)
|
2009-11-07 18:32:26 +01:00
|
|
|
|
toplevel-lambdas))
|
|
|
|
|
|
(else info))))
|
|
|
|
|
|
|
|
|
|
|
|
(let ((toplevel-calls (toplevel-procedure-calls info))
|
|
|
|
|
|
(lexical-lambdas (lexical-lambdas info))
|
|
|
|
|
|
(toplevel-lambdas (toplevel-lambdas info)))
|
|
|
|
|
|
|
|
|
|
|
|
(record-case x
|
|
|
|
|
|
((<toplevel-define> name exp)
|
|
|
|
|
|
(record-case exp
|
|
|
|
|
|
((<lambda> body)
|
|
|
|
|
|
(make-arity-info toplevel-calls
|
|
|
|
|
|
lexical-lambdas
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(vhash-consq name exp toplevel-lambdas)))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
((<toplevel-ref> name)
|
|
|
|
|
|
;; alias for another toplevel
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(let ((proc (vhash-assq name toplevel-lambdas)))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
(make-arity-info toplevel-calls
|
|
|
|
|
|
lexical-lambdas
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(vhash-consq (toplevel-define-name x)
|
|
|
|
|
|
(if (pair? proc)
|
|
|
|
|
|
(cdr proc)
|
|
|
|
|
|
exp)
|
|
|
|
|
|
toplevel-lambdas))))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
(else info)))
|
|
|
|
|
|
((<let> vars vals)
|
|
|
|
|
|
(fold extend info vars vals))
|
|
|
|
|
|
((<letrec> vars vals)
|
|
|
|
|
|
(fold extend info vars vals))
|
|
|
|
|
|
((<fix> vars vals)
|
|
|
|
|
|
(fold extend info vars vals))
|
|
|
|
|
|
|
|
|
|
|
|
((<application> proc args src)
|
|
|
|
|
|
(record-case proc
|
|
|
|
|
|
((<lambda> body)
|
|
|
|
|
|
(validate-arity proc x #t)
|
|
|
|
|
|
info)
|
|
|
|
|
|
((<toplevel-ref> name)
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(make-arity-info (vhash-consq name x toplevel-calls)
|
2009-11-07 18:32:26 +01:00
|
|
|
|
lexical-lambdas
|
|
|
|
|
|
toplevel-lambdas))
|
|
|
|
|
|
((<lexical-ref> gensym)
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(let ((proc (vhash-assq gensym lexical-lambdas)))
|
2009-11-07 18:32:26 +01:00
|
|
|
|
(if (pair? proc)
|
|
|
|
|
|
(record-case (cdr proc)
|
|
|
|
|
|
((<toplevel-ref> name)
|
|
|
|
|
|
;; alias to toplevel
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(make-arity-info (vhash-consq name x toplevel-calls)
|
2009-11-07 18:32:26 +01:00
|
|
|
|
lexical-lambdas
|
|
|
|
|
|
toplevel-lambdas))
|
|
|
|
|
|
(else
|
|
|
|
|
|
(validate-arity (cdr proc) x #t)
|
|
|
|
|
|
info))
|
|
|
|
|
|
|
|
|
|
|
|
;; If GENSYM wasn't found, it may be because it's an
|
|
|
|
|
|
;; argument of the procedure being compiled.
|
|
|
|
|
|
info)))
|
|
|
|
|
|
(else info)))
|
|
|
|
|
|
(else info))))
|
|
|
|
|
|
|
2010-01-08 12:02:00 +01:00
|
|
|
|
(lambda (x info env locs)
|
2009-11-07 18:32:26 +01:00
|
|
|
|
;; Up from X.
|
|
|
|
|
|
(define (shrink name val info)
|
|
|
|
|
|
;; Remove NAME from the lexical-lambdas of INFO.
|
|
|
|
|
|
(let ((toplevel-calls (toplevel-procedure-calls info))
|
|
|
|
|
|
(lexical-lambdas (lexical-lambdas info))
|
|
|
|
|
|
(toplevel-lambdas (toplevel-lambdas info)))
|
|
|
|
|
|
(make-arity-info toplevel-calls
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(if (vhash-assq name lexical-lambdas)
|
|
|
|
|
|
(vlist-tail lexical-lambdas)
|
|
|
|
|
|
lexical-lambdas)
|
2009-11-07 18:32:26 +01:00
|
|
|
|
toplevel-lambdas)))
|
|
|
|
|
|
|
|
|
|
|
|
(let ((toplevel-calls (toplevel-procedure-calls info))
|
|
|
|
|
|
(lexical-lambdas (lexical-lambdas info))
|
|
|
|
|
|
(toplevel-lambdas (toplevel-lambdas info)))
|
|
|
|
|
|
(record-case x
|
|
|
|
|
|
((<let> vars vals)
|
|
|
|
|
|
(fold shrink info vars vals))
|
|
|
|
|
|
((<letrec> vars vals)
|
|
|
|
|
|
(fold shrink info vars vals))
|
|
|
|
|
|
((<fix> vars vals)
|
|
|
|
|
|
(fold shrink info vars vals))
|
|
|
|
|
|
|
|
|
|
|
|
(else info))))
|
|
|
|
|
|
|
|
|
|
|
|
(lambda (result env)
|
|
|
|
|
|
;; Post-processing: check all top-level procedure calls that have been
|
|
|
|
|
|
;; encountered.
|
|
|
|
|
|
(let ((toplevel-calls (toplevel-procedure-calls result))
|
|
|
|
|
|
(toplevel-lambdas (toplevel-lambdas result)))
|
2010-02-03 00:00:05 +01:00
|
|
|
|
(vlist-for-each
|
|
|
|
|
|
(lambda (name+application)
|
|
|
|
|
|
(let* ((name (car name+application))
|
|
|
|
|
|
(application (cdr name+application))
|
|
|
|
|
|
(proc
|
|
|
|
|
|
(or (and=> (vhash-assq name toplevel-lambdas) cdr)
|
|
|
|
|
|
(and (module? env)
|
|
|
|
|
|
(false-if-exception
|
|
|
|
|
|
(module-ref env name)))))
|
|
|
|
|
|
(proc*
|
|
|
|
|
|
;; handle toplevel aliases
|
|
|
|
|
|
(if (toplevel-ref? proc)
|
|
|
|
|
|
(let ((name (toplevel-ref-name proc)))
|
|
|
|
|
|
(and (module? env)
|
|
|
|
|
|
(false-if-exception
|
|
|
|
|
|
(module-ref env name))))
|
|
|
|
|
|
proc)))
|
|
|
|
|
|
(if (or (lambda? proc*) (procedure? proc*))
|
|
|
|
|
|
(validate-arity proc* application (lambda? proc*)))))
|
|
|
|
|
|
toplevel-calls)))
|
|
|
|
|
|
|
|
|
|
|
|
(make-arity-info vlist-null vlist-null vlist-null)))
|