suzanne.soy/racket
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
15,034 Commits 17 Branches 54 Tags 278 MiB
249a9e38f7
Commit Graph

105 Commits

Author SHA1 Message Date
Stevie Strickland
290a73b56b Check and unwrap instead of grabbing the property value, then using it to 
unwrap.

svn: r18610
 2010-03-24 21:32:32 +00:00
Stevie Strickland
dde2011ec7 Add ->dm and case->m, which are the implicit method versions of ->d and 
case->.

This should be included in the release.

svn: r18579
 2010-03-18 21:28:41 +00:00
Stevie Strickland
5145638ac5 Since we're handling initialization and the result of this wrapper is a 
real (sub)class, we need this to be #f.

svn: r18544
 2010-03-15 09:19:50 +00:00
Stevie Strickland
924842d9e9 * Fix inits so that we either call the previous class/c wrapper's init 
or let continue-make-super take control if we were the first wrapper
* Fix up handling of init/contract sorting so that things are performed
  at the right phase (at least if we're going to treat names as symbols).
* Fix up new tests so that all contract layers are tested.

svn: r18540
 2010-03-15 08:29:22 +00:00
Stevie Strickland
7ec061cdbf * Factor out zipping of inits/contracts of the init closure 
* Add tests for higher-order behavior

svn: r18538
 2010-03-15 07:37:25 +00:00
Stevie Strickland
f9e404afbc * Fix bug in #:name for init contracts 
* Apply appropriate projections to init arguments.

svn: r18537
 2010-03-15 07:25:54 +00:00
Stevie Strickland
2f9717ca72 Start on the init function. First step -- just replace it if we have init 
contracts, but don't add any checking (yet).

svn: r18536
 2010-03-15 07:05:31 +00:00
Stevie Strickland
9640ea4e2c Sort the init contract forms. 
svn: r18535
 2010-03-15 07:03:54 +00:00
Stevie Strickland
1bd29dd7e9 Okay, we're going to add back init contracts, which I'd somewhat done before. 
In this version, we're not going to have any reasonable first-order checks as
to whether or not the class actually accepts the initialization arguments that
are being contracted.  I'm also just going to handle by-name at first, since
that's all I originally discussed in the paper.

svn: r18534
 2010-03-15 06:49:49 +00:00
Stevie Strickland
40b35a2b73 Now that we have a way of marking procedures as methods at runtime, 
fix up all projection applications and uncomment out the tests for
appropriate method arity errors.

svn: r18504
 2010-03-10 04:07:00 +00:00
Stevie Strickland
77f2555b9c Remove now-redundant object? check in is-a? 
svn: r18448
 2010-03-03 17:04:41 +00:00
Stevie Strickland
3fb2e1d9ae Check that we have an object before trying to unwrap it in class-field-*. 
svn: r18443
 2010-03-03 14:39:44 +00:00
Stevie Strickland
16dbb0edc1 Fix bug in field mutation on contracted objects. 
svn: r18390
 2010-02-27 20:59:40 +00:00
Sam Tobin-Hochstadt
7accc0bd7d Need to unwrap objects at two other points: 
* When the C++ code uses direct access on objects (look for check_valid)
 * In the functions returned by class-field-accessor/mutator

svn: r18344
 2010-02-25 23:42:13 +00:00
Stevie Strickland
1f10c4e422 Have external field accessors/mutators just redirect to the old object 
themselves.  Also, have all objects carry the prop:unwrap property.

svn: r18330
 2010-02-24 23:32:44 +00:00
Stevie Strickland
af0387e3a6 Fuck you, C, and the flea-bitten, mange-covered, syphilis-ridden horse you 
rode in on.

svn: r18318
 2010-02-24 16:35:30 +00:00
Stevie Strickland
60b6c81f9f Okay, updates to all this crapola. Going to try and see what happens if I 
only attach prop:unwrap to wrapped objects.

svn: r18313
 2010-02-23 23:56:08 +00:00
Stevie Strickland
f60645da2c ARGHGHGHGHGHGHGHGGHGH 
svn: r18308
 2010-02-23 21:43:08 +00:00
Stevie Strickland
e1cd160a40 Now that the C code is back at a compile-ready point, I'll check in. 
svn: r18306
 2010-02-23 20:57:22 +00:00
Stevie Strickland
472ef1e873 Other cleanups. 
svn: r18303
 2010-02-23 15:59:22 +00:00
Stevie Strickland
1eeb27fc30 Small fixes in make-wrapper-class/object. 
svn: r18302
 2010-02-23 15:58:00 +00:00
Stevie Strickland
6583b0b77c My first foray into the C side of things. 
svn: r18299
 2010-02-23 14:55:24 +00:00
Stevie Strickland
09425bc801 Keep the original class in the supers list. Also, copy over the no-super-init? 
flag.

svn: r18296
 2010-02-23 12:51:27 +00:00
Stevie Strickland
e4f7f0032e Get rid of the loop that's no longer a loop, and also add in the necessary 
object unwrapping.

svn: r18288
 2010-02-23 04:13:09 +00:00
Stevie Strickland
978a9586f5 We no longer need the #:error thing here, because we've fixed object-contract 
for real now.

svn: r18286
 2010-02-23 04:02:03 +00:00
Stevie Strickland
14ab0175c3 Okay, expanding field accesses and mutations to basically inline the 
unwrapping operation helps a bit, especially with inherited fields.
Unfortunately, as one might expect, TANSTAAFL applies here.  In order
to make sure that we keep the contracted objects around as much as
possible to make sure there are no holes, we end up making local and
inherited field access codes 2-3x more than they did before.  However,
this is still something on the order of 5x faster than external
access.  But blah.

CONTRACTS ARE NOT FREE.  Just ask your local lawyer.

svn: r18285
 2010-02-23 03:15:43 +00:00
Stevie Strickland
53381bbf03 Remove unwrapping in find-method/who until I figure out what I actually need 
to do.

Also fix up is-a? and subclass? so that they should work the same when
contracts are removed from a program.

svn: r18282
 2010-02-23 01:15:11 +00:00
Stevie Strickland
f1b0bfdd79 Yeah, accessors need arguments. 
svn: r18281
 2010-02-23 00:46:47 +00:00
Stevie Strickland
cfdb9dd39b Time to unveil object/c. 
svn: r18280
 2010-02-23 00:43:25 +00:00
Stevie Strickland
ab2561e08a Now we don't need to recur down to unwrap something, but if we get a wrapped 
primitive object in a method send, we need to unwrap all objects for its
method.

svn: r18279
 2010-02-23 00:40:59 +00:00
Stevie Strickland
2af44afb17 Now I see -- I was handling local fields in an incorrect manner. We don't 
want later projections to affect local accesses or mutations -- so we just
have to add the unwrap check in case it's a wrapped object.

svn: r18277
 2010-02-22 22:43:47 +00:00
Stevie Strickland
0e3af71176 So now object-contract works again, but we seem to have introduced a bug 
in the class/c inherit-field form, so now time to fix that.

svn: r18276
 2010-02-22 22:26:27 +00:00
Stevie Strickland
a3a1d0d9c7 Fix shadowing in make-wrapper-class. Now delay lookup for 
accessors/mutators used for internal field access.  This fixes public
fields, but not private fields.  Next should fix that up.

Will definitely need to benchmark all this delay though.

svn: r18275
 2010-02-22 22:10:30 +00:00
Stevie Strickland
d820493feb First cut of converting object-contract to share a common base that 
object/c will also use.

svn: r18274
 2010-02-22 21:55:32 +00:00
Stevie Strickland
815dd80923 Sync up to catch my fix. 
svn: r18270
 2010-02-22 20:58:53 +00:00
Stevie Strickland
11b8fd4204 Fix vector creation for internal field access. 
svn: r18269
 2010-02-22 20:57:36 +00:00
Stevie Strickland
c2fcdbba65 Class Contracts Phase 2: Object/c Boogaloo 
This isn't just a copy of trunk r18264 -- it has a slight difference in how
local field accessors and mutators are handled that will eventually play a
larger role.

svn: r18265
 2010-02-22 19:09:42 +00:00
Stevie Strickland
a0769da5ea Add the contract shorthands for -> and ->* to use for methods where we don't 
care about properties of this.

svn: r18248
 2010-02-21 02:54:06 +00:00
Stevie Strickland
ffa34e1f7d Add augride, which is like augment but enables the contract writer to give 
subclasses an idea of whether a method can be augmented (augment) or whether
a method augmentation can be overridden (augride).

svn: r18240
 2010-02-21 00:17:42 +00:00
Stevie Strickland
f72ca7bb1b Now inherit works (and tests!) 
svn: r18237
 2010-02-20 22:54:11 +00:00
Stevie Strickland
370792b881 Refactoring done, and I think that's actually cleaned up things a bit. Now 
to handle inherit.

svn: r18236
 2010-02-20 22:44:53 +00:00
Stevie Strickland
b589c3c230 More preparation to move all the int-method/dynamic-proj expansion into 
class/c-proj instead of compose-class.

svn: r18235
 2010-02-20 22:00:45 +00:00
Stevie Strickland
66ce493ede Adding original class field (we'll see what this is for in a sec.) 
svn: r18234
 2010-02-20 21:48:00 +00:00
Stevie Strickland
a0fdeff509 First order checks. 
svn: r18233
 2010-02-20 21:34:57 +00:00
Stevie Strickland
a4d6252d16 Start inherit contracts (which are useful for mixins). Tests, plus parsing. 
svn: r18232
 2010-02-20 21:28:20 +00:00
Stevie Strickland
cd4aa4c6f6 Forgot about the no-contract forms, so needed to add tests for those, also. 
svn: r18218
 2010-02-20 10:40:50 +00:00
Stevie Strickland
7830d55b42 Okay, that does it for augment, which means I'm done with coding. Now just 
documentation and benchmarking, then this can go on trunk.

svn: r18217
 2010-02-20 10:09:37 +00:00
Stevie Strickland
67d47e0a1d Fixes in override ctcs and test suite. I thought I ran it, so I find it 
weird that I found these on a subsequent run when adding some quick augment
tests to start the next batch.  (Oh, those are included also.)

svn: r18215
 2010-02-20 09:40:41 +00:00
Stevie Strickland
b5e2d5f93e Okay, now override contracts are done, so only augments remain. 
svn: r18214
 2010-02-20 09:30:40 +00:00
Stevie Strickland
3c1004fd05 Okay, we should be fixed up in compose-class, now we just need to start 
handling the projections in class/c-proj.

svn: r18213
 2010-02-20 09:14:14 +00:00