Jean Privat [Thu, 16 Apr 2015 13:34:47 +0000 (20:34 +0700)]
tests: update test_ffi_c_operators for << and >>
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 16 Apr 2015 11:04:11 +0000 (18:04 +0700)]
tests: add error_operators.nit
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 16 Apr 2015 11:01:11 +0000 (18:01 +0700)]
modelize_property: implement special validity rules for operators
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 16 Apr 2015 04:13:24 +0000 (11:13 +0700)]
Merge: Extends operators
* Add missing combined assignments: `*=`, `/=`, `%=`, `**=`, `<<=`, `>>=`
* Add missing unary + operator (why?)
* Keep operators in the AST for better messages
* Factorize some information in the AST to simplify code in other modules
Pull-Request: #1263
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Jean Privat [Wed, 15 Apr 2015 05:40:32 +0000 (12:40 +0700)]
tests: update other sav for new AST of location in error messages
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 05:39:56 +0000 (12:39 +0700)]
tests: extends base_combined_assignment and test_ffi_c_operators
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 05:36:24 +0000 (12:36 +0700)]
lib/stc/kernel: add << and >> in Int
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 05:37:17 +0000 (12:37 +0700)]
nitpretty: handle AUplusExpr and simplify binary operations
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 05:38:55 +0000 (12:38 +0700)]
typing: use new `operator` and `n_op` from ABinopExpr and AAssignOp
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 05:00:13 +0000 (12:00 +0700)]
src: implement unary plus
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 04:50:36 +0000 (11:50 +0700)]
parser: regenerate
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 03:50:33 +0000 (10:50 +0700)]
grammar: add more combined assignment operator
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 04:49:59 +0000 (11:49 +0700)]
grammar: introduce n_op and operator in AAssignOp
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 04:48:42 +0000 (11:48 +0700)]
parser: add ABinopExpr::operator to avoid refining each operator class in other modules
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 04:09:18 +0000 (11:09 +0700)]
grammar: keep operators in the AST for better messages
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 03:53:38 +0000 (10:53 +0700)]
grammar: as unary plus operation (only for thr sake on completion)
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 03:34:16 +0000 (10:34 +0700)]
Merge: Review errors messages
A big PR but that just review the errors and warning messages printed by the tools.
I tries to follow a standard format:
eg. "Error: cannot find method `foo`."
* start with a message type like "Error:", "Syntax Error:", "Warning:".
The type is capitalized and terminated by a column.
The rest on the message starts with a lowercase letter and is terminated with a dot.
* be short and fit on a single line.
* have meaningful information first.
This helps the reader and remain usable
when truncated, by an IDE for instance.
* enclose identifiers, keywords and pieces of code with back-quotes.
Do not hesitate no nitpick phrasing or wording or propose better messages.
Close #931
Pull-Request: #1253
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Jean Privat [Wed, 15 Apr 2015 03:34:05 +0000 (10:34 +0700)]
Merge: Check contrib
Add a script `misc/jenkins/check_contrib.sh` to automatically compile and tests things in contrib and examples.
New (and existing) programs in these directory should try to add some basic (or extensive) checking.
Note that I used the name `check` that is the standard name for Makefiles (`tests` seems to be an invention of myself) http://www.gnu.org/software/make/manual/make.html#Standard-Targets
Pull-Request: #1259
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Jean Privat [Wed, 15 Apr 2015 03:33:59 +0000 (10:33 +0700)]
Merge: tests: add --autosav to automatically update the sav directory.
Advantages:
* avoid manual error-prone update (especially on large tests)
* diffs on the sav/ files may be easier to see with git tools
Drawbacks:
* the wrong sav dir can by used for specific engines/uname/etc.
* manual inspection still required
Is used for #1253 because manual updates are a PITA
Pull-Request: #1258
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Jean Privat [Fri, 10 Apr 2015 15:44:46 +0000 (22:44 +0700)]
tests: update sav with new error messages
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 15 Apr 2015 01:14:56 +0000 (08:14 +0700)]
src: improve messages (and sometime location) of errors and warnings
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Tue, 14 Apr 2015 10:07:29 +0000 (17:07 +0700)]
model: add `MClass::signature_to_s` to use in messages.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 9 Apr 2015 12:40:28 +0000 (19:40 +0700)]
typing: identify the node associated to method name
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Tue, 14 Apr 2015 11:57:37 +0000 (18:57 +0700)]
contrib: add some check rules
`check` is the expected name for Makefile, `tests` is a nitism
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Tue, 14 Apr 2015 11:56:44 +0000 (18:56 +0700)]
jenkins: add check_contrib.sh to automatically compile and check contribs
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Tue, 14 Apr 2015 11:11:11 +0000 (18:11 +0700)]
tests: add --autosav to automatically update the sav directory.
Advantages:
* avoid manual error-prone update (especially on large tests)
* diffs on the sav/ files may be easier to see with git tools
Drawbacks:
* the wrong sav dir can by used for specific engines/uname/etc.
* manual inspection still required
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Tue, 14 Apr 2015 09:33:40 +0000 (16:33 +0700)]
Merge: Fix type adaptation when there is loops
Basically one of the oldest bug of Nit.
The point of adaptive typing is that the static types of variables follows the static types of the assigned values, type tests and comparison to null.
This is a great feature of the language but was buggy because loops where not taken in account (old TODO), until the present PR.
The basic idea is just to track if a type adaptation occurs, if yes replay the whole typing phase.
The immediate drawback is that now typing require 2 passes most of the time, and sometime more.
This PR was a roller-coaster of emotions to develop since a lot a surprise issues were discovered.
* bug in flow that give me hours of despair.
* duplication of error messages; because multiple passes.
* premature warning/error messages; because some error conditions on types are detected at the first pass but a second pass find it is fine in fact... but too late the error message was printed.
* almost work at the first time (modulo the previous issues)
* only two bugs found in the whole code, quite impressive, and one of them is not a real bug but an autocast issue.
The issue about error duplication and premature errors is only workadounded (or fixme) in most places.
But proper solutions will be required in future PR.
close #647 and close #86, then reopen it again, then reclose it.
Pull-Request: #1257
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Tue, 14 Apr 2015 09:33:32 +0000 (16:33 +0700)]
Merge: Intern new
A small short PR that enables a `new`-factory to instantiate itself.
previously there was no way to do that. The following code will stack-overflow:
~~~nit
class A
new do
var res = new A
res.do_thigns
return res
end
end
var a = new A # infinite recursion of `new A`
~~~
Note: this is a very bad example as what is done in the previous `new` should be done in an `init` instead since it is related to the initialization of the object. A `new` factory should be exclusively used for some factory-concern like returning a more specialized object or returning an already existing object.
With this PR, the primitive constructor is available and is called `intern`, as the annotation that indicates things that cannot be programmed in Nit but are provided by the execution engine.
So in the previous example, just use this primitive constructor instead of doing a recursive call:
~~~
var res = new A.intern
~~~
This intern constructor just do the allocation and the initialization of attributes with default values.
`init` is not called.
Maybe it should not do the initialization of attributes neither but if I skip them, there is no way for the programmer to ask for them manually. whereas `init` in not automatically called but can be called manually.
note: this PR is a small step toward the conclusion of the constructor saga, more will come soon (I hope)
Pull-Request: #1252
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Jean Privat [Tue, 14 Apr 2015 09:33:22 +0000 (16:33 +0700)]
Merge: Improved parallelization phase
The "threaded" annotation now supports functions with parameters !
Still need to check for eventual self calls in the body of the annotated function
Pull-Request: #1255
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Reviewed-by: Jean Privat <jean@pryen.org>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Jean Privat [Tue, 14 Apr 2015 09:32:55 +0000 (16:32 +0700)]
Merge: Make the generated iOS project compilable in the XCode GUI
I originally incorporated the test project in the generated XCode workspace thinking that we could use it somehow. But it duplicates services from nitunits and it would be much more complex to use.
I remove the test project now because it is incomplete and raise errors from the XCode GUI.
Pull-Request: #1254
Reviewed-by: ArthurDelamare <arthur.delamare@viacesi.fr>
Reviewed-by: Jean Privat <jean@pryen.org>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Jean Privat [Tue, 14 Apr 2015 09:32:40 +0000 (16:32 +0700)]
Merge: Vim plugin fixes and make Nitdoc a command with search capabilities
You can now use Nitdoc as a command in Vim.
* Search the doc for the word under the cursor with `:Nitdoc`
* Display the doc of the `String` class (followed by all docs using the word "String", case-insensitive) with `:Nitdoc String`
* Display the doc of `Int::%` with `:Nitdoc modulo`
* Search for a string of words (in the given order) with `:Nitdoc not null`
Some query can produce quite a lot results. `:Nitdoc the` will fill the preview window with 33 379 lines of doc, but there no noticeable delay.
### Bonus tip!
Navigate in the doc by moving the cursor above a word in the preview window and using ctrl-D (or `:Nitdoc`), exactly as you would do in Nit code.
Pull-Request: #1256
Reviewed-by: Jean Privat <jean@pryen.org>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Sun, 12 Apr 2015 04:14:58 +0000 (11:14 +0700)]
tests: add base_adaptive_loop*.nit
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:14:40 +0000 (00:14 +0700)]
tests: update sav for existing tests
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:13:56 +0000 (00:13 +0700)]
toolcontext: avoid duplicated messages
This is just a workaround to limit repeated errors message because of
adapting typing
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:12:50 +0000 (00:12 +0700)]
lib/ai/backtrack: add a missing assert
or else the possible null will loop and cause errors.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:11:40 +0000 (00:11 +0700)]
lib/a_star: avoid bad autocast
Seriously, this one case might be a strong argument against autocasts
~~~
frontier_node = frontier_node.best_source.as(not null)
~~~
is auto-casted as
~~~
frontier_node = frontier_node.best_source.as(not null).as(N)
~~~
bun since the N is `nullable Node`, the not-null information was lost in the
cast.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sun, 12 Apr 2015 04:13:24 +0000 (11:13 +0700)]
compiler: consider untyped expressions and statement as dead code
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:09:04 +0000 (00:09 +0700)]
typing: handle `a == null` when `a` is null
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:07:41 +0000 (00:07 +0700)]
typing: shortcut `get_variable` on never-adapted variables
Signed-off-by: Jean Privat <jean@pryen.org>
Alexis Laferrière [Mon, 13 Apr 2015 03:52:34 +0000 (23:52 -0400)]
misc/vim: inform the user when no results are found
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Alexis Laferrière [Sat, 11 Apr 2015 14:21:51 +0000 (10:21 -0400)]
misc/vim: update README file
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Jean Privat [Sat, 11 Apr 2015 17:06:45 +0000 (00:06 +0700)]
typing: ensure monotony on literal arrays
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:06:06 +0000 (00:06 +0700)]
typing: revisit on type-adaptation, until there is no more adaptation
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:04:16 +0000 (00:04 +0700)]
typing: mark variable dirty on type_adaptation
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 17:00:09 +0000 (00:00 +0700)]
typing: compute FlowContext::collect_types without recursion
So this avoid infinite recursion on loops
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 16:55:07 +0000 (23:55 +0700)]
typing: remove cache in FlowContext
Because type adaptation require recomputation on loops
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 16:52:46 +0000 (23:52 +0700)]
flow: improve dot-rendering for debugging
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Sat, 11 Apr 2015 16:52:01 +0000 (23:52 +0700)]
flow: fix loop links on continue
Targets the begin of the block
Signed-off-by: Jean Privat <jean@pryen.org>
Alexis Laferrière [Sat, 11 Apr 2015 14:32:45 +0000 (10:32 -0400)]
misc/vim: delete the first empty line of the preview window on Nitdoc command
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Alexis Laferrière [Sat, 11 Apr 2015 14:12:30 +0000 (10:12 -0400)]
misc/vim: the Nitdoc command search with priorities
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Alexis Laferrière [Sat, 11 Apr 2015 13:41:21 +0000 (09:41 -0400)]
misc/vim: make Nitdoc a command
Use this vim command to show the doc of any entities `:Nitdoc Object`.
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Alexis Laferrière [Sat, 11 Apr 2015 14:11:00 +0000 (10:11 -0400)]
misc/vim: fix omnifunc order of results found in the long doc
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Alexis Laferrière [Sat, 11 Apr 2015 14:10:15 +0000 (10:10 -0400)]
misc/vim: make private functions local to the script
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Alexis Laferrière [Fri, 10 Apr 2015 19:09:29 +0000 (15:09 -0400)]
nitc/ios: use a camel case version of app namespace for some version of XCode
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Alexis Laferrière [Fri, 10 Apr 2015 17:21:10 +0000 (13:21 -0400)]
nitc/ios: remove all testing related entities from the xcode project
Signed-off-by: Alexis Laferrière <alexis.laf@xymus.net>
Jean Privat [Fri, 10 Apr 2015 15:19:40 +0000 (22:19 +0700)]
Merge: Sys is top
And when I say *top* I mean *great*. But I also mean *top* in fact.
The idea is to move the top-level methods (those defined outside classes) from Object to Sys.
While this is a cosmetic move, it has a lot of benefits:
* top-level methods get a real meaningful receiver: the current system. It is meaningful both on the declaration side and the call side.
* no more need to distinguish the concept of top-level methods with their rules and semantic, so this simplify the language with one less thing (almost, see bellow)
* self is now usable in top-level methods, not that useful now because it is `sys`, a singleton, but this allow some kind of inheritance if you add super-classes to Sys.
* no more name conflicts between a standard class method and a top-level method
* no more useless slots in table of classes for crazy methods defined in Object and never redefined
* close #461 and close #1081 by making them irrelevant
* Let us see a bright future where the singleton Sys become a multiton and allow specific isolation of computation. Specific Sys will then be active objects (threads, computation node, actors, whatever) isolated with their own specific cloud of objects and efficient lock-less concurrency, dedicated memory model (realtime?), and why not transparent distribution
There is still two drawbacks
* what is the status of `sys` that represent the current `Sys`? I hard-coded it to stay a method in Object. One way to solve the probem is to make it a keyword (like `self`)
* when doing `foo(x)`, first this tries `self.foo(x)` then `sys.foo(x)` thus this is some additional rule of the language.
Pull-Request: #1249
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Jean Privat [Fri, 10 Apr 2015 11:27:53 +0000 (18:27 +0700)]
Merge: Robust IO
Some cleanup on lib/standard/file.nit
Especially close #755
There is still some work to do and question to answer from an API point of view.
Pull-Request: #1246
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Romain Chanoir <chanoir.romain@courrier.uqam.ca>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Fri, 10 Apr 2015 11:27:45 +0000 (18:27 +0700)]
Merge: miniclean quadtrees
Still a lot of warning remains...
Pull-Request: #1251
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Romain Chanoir <chanoir.romain@courrier.uqam.ca>
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Jean Privat [Fri, 10 Apr 2015 10:31:17 +0000 (17:31 +0700)]
tests: remove fixme sav/nitg-g/fixme/test_deriving_alt1.res
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 13:53:37 +0000 (20:53 +0700)]
tests: ffi use `Sys_` instead of `Object_` prefix
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 13:53:03 +0000 (20:53 +0700)]
tests: update tests results related to top-level or Object->Sys
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 15:51:44 +0000 (22:51 +0700)]
code: remove `protected` from top-level methods (now in sys)
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 13:51:31 +0000 (20:51 +0700)]
typing: remove most top-level things
only kept for `new` factories and `sys` (for the moment)
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 13:50:04 +0000 (20:50 +0700)]
lib/md5: do not import `print` in FFI
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 13:15:03 +0000 (20:15 +0700)]
mixin: look for top-level methods in `Sys`.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 13:14:19 +0000 (20:14 +0700)]
modelize_classes: top-level methods are defined in Sys, not Object
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 13:11:35 +0000 (20:11 +0700)]
engines: implement special fall-back to sys
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 13:10:54 +0000 (20:10 +0700)]
typing: look for method in `sys` when they are not in `self`
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 10:18:00 +0000 (17:18 +0700)]
modelize classes: process AStdClassdef before
So that non-AStdClassdef classes can be attached to existing ones if any
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 10 Apr 2015 10:09:18 +0000 (17:09 +0700)]
niti: implements native `chdir` and `mkdir` with a return value
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 10 Apr 2015 05:20:58 +0000 (12:20 +0700)]
tests: add base_new_intern.nit
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 10 Apr 2015 05:16:03 +0000 (12:16 +0700)]
typing: `intern` is used to refer to the plain vanilla constructor
It is used to implement specific factories inside classes
~~~
class A
new do
var res = new A.intern
res.do_things
return res
end
end
~~~
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Fri, 10 Apr 2015 05:13:39 +0000 (12:13 +0700)]
engines: ANewExpr just return the plain instance if no callsite
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 9 Apr 2015 14:50:55 +0000 (21:50 +0700)]
tests: update sav related to ligne changes in kernel
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 9 Apr 2015 14:50:27 +0000 (21:50 +0700)]
contrib/nitester: no more free to call
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 6 Apr 2015 15:24:20 +0000 (22:24 +0700)]
lib: move errno and strerror to legacy FFI
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 6 Apr 2015 15:10:17 +0000 (22:10 +0700)]
lib/file: once some global constants
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 6 Apr 2015 15:09:55 +0000 (22:09 +0700)]
lib/file: some methods return a nullable Error on error
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 6 Apr 2015 15:08:52 +0000 (22:08 +0700)]
lib/file: String::files does not exist on error
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 6 Apr 2015 15:07:30 +0000 (22:07 +0700)]
lib/file: remove useless protected for top-level methods
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Mon, 6 Apr 2015 15:06:59 +0000 (22:06 +0700)]
lib/file: expose FileStat instead of NativeFileStat
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 9 Apr 2015 13:30:54 +0000 (20:30 +0700)]
quatree: cleaup indentation
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 9 Apr 2015 13:30:39 +0000 (20:30 +0700)]
quadtree: use new constructors
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Thu, 9 Apr 2015 05:24:43 +0000 (12:24 +0700)]
Merge: Misc AST
Various small improvements on the ast side.
The main improvement is the lower consumption of Location instances.
Before: 681k where created; for 174MIr
After: 472k are created (-30%); for 120MIr (-31%)
However, the final impact should be negligible.
Other changes are related to error messages in order to have them more useful in other context.
As a proof of concept, nitlight can show some warnings and errors in the generated HTML.
There is still issue, especially: how to display them property. Currently they are in a popupable dropdown boxes, but the with is too small and nested info boxes show unreadable stacked things.
Eg, for `return new Foo[nullable Bar]`, how can I show
* information on the classes Foo and Bar
* error on the whole `nullable Bar` because Foo is in fact not generic
* error on the whole `new Foo[nullable Bar]` because in fact we are in a procedure, so no return value is expected
Pull-Request: #1248
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Thu, 9 Apr 2015 05:24:37 +0000 (12:24 +0700)]
Merge: new `with` statement
Introduce the `with` statement that is analogous to the `with` of Python or the `uning` of C#.
~~~nit
with x = something do
x.work
end
~~~
That is equivalent with
~~~nit
do
var x = something
x.start
x.work
x.finish
end
~~~
Note that an alternative syntax, without the local variable exists:
~~~nit
with something do
work
end
~~~
that is used when we want some automatic control or whatever. Eg resource allocation or critical section.
~~~nit
do
var tmp = something
tmp.start
work
tmp.stop
end
~~~
Most real-world examples could be
~~~nit
with file = path.create do
file.write("Bla")
end
~~~
and
~~~nit
with a_mutex do
work
end
~~~
## Names
I reused the `finish` method from Iterator.
I introduced the antonym `start` for the other side.
Note that the `start` is important because it helps the object to know it has to prepare itself for a `finish`.
I did not introduce an interface, mainly because I have no idea for a name. C# names it `IDisposable` and the `finish` method is called `Dispose`. There is no `start` because C# designers are lunatic sometime.
In python there is no interface (obviously) and the method are called
`__enter__` and `__exit__` because Python likes identifiers that resemble words eaten by boas.
Note that the returned variable is not the original expression but the result of `__enter__`.
With the Python semantic, but the Nit syntax, the first example will be equivalent to
~~~nit
do
var tmp = something
var x = tmp.start
x.work
x.finish
end
~~~
I am not sure of the benefits of the additional complexity, but the the more I thing about it the more I think this could be useful, ex for the control case the `start` can return the handler to `finish`.
The issue is that proposed syntax: `with x = something` does not make sense (because `tmp=something` and `x=tmp.start`) thus should be replaced with something else; Python uses `with something as x:` that we can reuse but I do not want because it is ugly.
An other alternative I tough of was to reuse the `for` keyword. but It is convoluted.
The two previous real world examples will then be
~~~nit
for file = path.create do
file.write("Bla")
end
~~~
and
~~~nit
for a_mutex do
work
end
~~~
Pull-Request: #1243
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Reviewed-by: Etienne M. Gagnon <egagnon@j-meg.com>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Jean Privat [Wed, 8 Apr 2015 14:39:22 +0000 (21:39 +0700)]
tests: update nitlight and test_parser
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 14:36:59 +0000 (21:36 +0700)]
hightlight: remove two asserts that make #1247 crashes
the result is still buggy but things keep-going
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 05:29:24 +0000 (12:29 +0700)]
test_parser: add option `-x` to output XML
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 05:22:11 +0000 (12:22 +0700)]
highlight: add infoboxes with messages
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 05:21:31 +0000 (12:21 +0700)]
highlight: factorize the creation of tags
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 05:08:41 +0000 (12:08 +0700)]
toolcontext: error methods return the messages, to add information
unused yet.
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 16:10:42 +0000 (23:10 +0700)]
toolcontext: attach errors message to their location
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 05:01:07 +0000 (12:01 +0700)]
parser: reuse child location when possible
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 05:00:20 +0000 (12:00 +0700)]
lexer: do not create useless location (for whitespaces)
Signed-off-by: Jean Privat <jean@pryen.org>
Jean Privat [Wed, 8 Apr 2015 01:03:25 +0000 (08:03 +0700)]
Merge: Parallelization phase introduction
Introduces the begining of the parallelization in Nit !
For now, only accepts the "threaded" annotation on methods without return value and without parameters.
Coming next :
Parametrized methods,
methods with return value,
for loops.
Pull-Request: #1247
Reviewed-by: Alexis Laferrière <alexis.laf@xymus.net>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Jean Privat <jean@pryen.org>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Wed, 8 Apr 2015 01:01:49 +0000 (08:01 +0700)]
Merge: Not null types
In order to fix #86 and #1238 some preliminary work to solve remaining issues with the type system is needed. This PR is a step toward this goal.
This introduce not-null types, that is a modifier to indicate that a type cannot contain null.
Basically, this new modifier is almost useless because it is the semantic of all the types (except obviously null and nullable things). Except in one case: when one adapts a formal type whose bound is nullable.
Before the PR, the semantic was the following
~~~nit
class A[E: nullable Object]
fun foo(e: E, o: nullable Object) do
var o2 = o.as(not null) # the static type of o2 is `Object`
print o2 # OK
var e2 = e.as(not null) # the static type of e2 is still `E` because there is no `nullable` to remove
print e2 # Error: expected Object, got E
end
end
~~~
Obviously, the issue was not that important because people managed to program complex things in Nit and I do not remember getting some complain about that particular issue. For the rare cases of this unexpected behavior, a workaround was possible: to cast on the non-nullable bound
~~~nit
var e2 = e.as(Object)
print e2 # OK
~~~
Nevertheless, the behavior was still buggy since type information was lost and not POLA. Moreover, `!= null` and `or else` did not have a workaround.
So, this PR introduces a special new type-modifier for this case so that everything become sensible.
~~~nit
var e2 = e.as(not null) # the static type of e2 is now `not null E`
print e2 # OK
~~~
Moreover, a lot of local refactorisation was done in model.nit and typing.nit to clean and harmonize the code. So that independently of the new notnull modifier, the code is cleaner, some bugs where removed and some small features added, especially the detection of useless `or else`.
Last, but not least, the `not null` thing is only an internal modifier and is not usable as a syntactic type construction (the grammar and the AST is unchanged); `not null` can however be shown to the programmer in messages.
~~~nit
var e2 = e.as(not null) # the static type of e2 is now `not null E`
var e3 = e2.as(not null) # << Warning: expression is not null, since it is a `not null E` >>
~~~
I could easily add `not null` as a specific syntactic construction since everything internally is ready. but 1. does this worth it?. 2. I do not want to conflict with #1243 that also change the grammar.
As an example, is it useful to write the following? (currently refused but very easy to add after this PR)
~~~nit
class A[E: nullable Object]
fun foo(e: not null E): not null E do
var x = e.to_s # no null pointer exception
# ...
return e
end
end
var a = new A[nullable Int]
var i = a.foo(5)
~~~
Pull-Request: #1244
Reviewed-by: Etienne M. Gagnon <egagnon@j-meg.com>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Romain Chanoir <chanoir.romain@courrier.uqam.ca>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Wed, 8 Apr 2015 01:01:15 +0000 (08:01 +0700)]
Merge: Improve checking of virtual types
This PR does 2 things to avoid runtime errors during nitc. Invalid programs that crashed the compiler now display errors messages.
* improve the static detection of loops in virtual types.
* consider types in the signatures of properties to be fragile since they contains potentially bad virtual types.
If merged with #1241 then nitpick can process the whole tests directory without crashing (impressive, since there is 1584 border-case files).
~~~
$ ./nitpick ../tests/*.nit ../tests/alt/*.nit -I ../lib/standard/ -I ../lib/standard/collection/
[... lots of errors ...]
Errors: 1766. Warnings: 195.
~~~
Pull-Request: #1245
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Romain Chanoir <chanoir.romain@courrier.uqam.ca>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Wed, 8 Apr 2015 01:01:07 +0000 (08:01 +0700)]
Merge: Clean benches
Some janitoring in the benchmarks directory
Pull-Request: #1242
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Jean Privat [Wed, 8 Apr 2015 01:00:53 +0000 (08:00 +0700)]
Merge: Robust keep-going
Detection and resolution of crashes or irrelevant error messages with tools that keeps going (eg. nitpick).
Historically, the compiler just exited at the end on a step when an error occured. The ToolContext#keep_going flag disable this automatic exit and was introduced for tool that are expected to be launched on broken code: nitpick and nitlight. However most parts of the code expected that:
* the previous steps where completed successfully, and expected information where available and correct.
* they are that special important snowflake can just exit the whole program when they are unhappy.
These rules can make sense for a compiler but are to strict for most SE tools that need to work (or continue working) when some random module contain an error.
Thus this PR improve the robustness of various parts of the code for the keep-going case so that we do not exit or crash before having processed all the modules.
With this PR, the following command does not crash anymore:
~~~
$ nitpick `nitls -rps ../lib ../contrib ../examples`
[...]
Errors: 10. Warnings: 31.
~~~
Note that `nitpick ../tests` still crashes. There is ugly errors hard to recover from in this directory.
Maybe, one day most of the tools can be made keep going by default.
Pull-Request: #1241
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>
Reviewed-by: Etienne M. Gagnon <egagnon@j-meg.com>
Jean Privat [Tue, 7 Apr 2015 10:37:36 +0000 (17:37 +0700)]
src/hightlight: hightlight and do not crash on MNotNullType
Also add MNullType that where missing.
Signed-off-by: Jean Privat <jean@pryen.org>