Support JSON Schema properties with values in the JSON Schema Compiler.

Support JSON Schema properties with values in the JSON Schema Compiler.

This brings the Extension IDL Schema feature set closer to the JSON Schema
feature set. In particular, the MDNS API needs this functionality.

Unfortunately the PPAPI IDL parser doesn't support WebIDL's "const" keyword,
which would be the right way to do this. Instead we fake it (though, consistent
with functions and events); to define a JSON Schema property, IDL Schemas
declare a 'Properties' namespace populated with no-argument functions,
annotating each function with the value they should be mapped to.

For example, this IDL:

namespace myApi {
  interface Properties {
    // A constant.
    [value=42] static long someConstant();
  };
};

will generate this JSON Schema:

{
  "namespace": "myApi",
  "properties": {
    "someConstant": {
      "value": "42",
      "description": "A constant",
      "type": "integer",
    }
  }
}

and finally, an integer myApi::some_constant equal to 42 in C++.

R=rdevlin.cronin@chromium.org

Committed: https://crrev.com/d94f954f5e1c32e7be85e2ba835a143fd388a0d8
Cr-Commit-Position: refs/heads/master@{#323401}

[not at google - send to devlin, 2015-03-31 22:11:09]

How's your python?

[Devlin, 2015-04-01 22:46:20]

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
File tools/json_schema_compiler/idl_schema.py (right):

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:208: 'with a non-void return type')
including "name" here would probably be helpful if this fails.

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:234: properties['enum'] =
[self.cast_from_json_type(properties['type'], enum)
why not:
map(lambda e: self.cast_from_json_type(properties['type'], e), enum_values)?

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:411: # it as a list. Convert back to
key-value pairs.
Nit: put this comment two lines lower (above where you do the conversion).

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:414: for prop in properties_as_list:
Would this be crazy?
map(lambda p: self.properties[p.pop('name')] = p, properties_as_list)

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
File tools/json_schema_compiler/test/idl_properties.idl (right):

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/test/idl_properties.idl:17: // Unvalued property.
Why do we want to allow this?

[Red Daly, 2015-04-01 23:12:54]

reddaly@chromium.org changed reviewers:
+ reddaly@chromium.org

[Red Daly, 2015-04-01 23:12:54]

Thanks for implementing this!

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
File tools/json_schema_compiler/idl_schema.py (right):

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:416: self.properties[name] = prop
Does the IDL parser ensure there are no duplicate names?

[not at google - send to devlin, 2015-04-01 23:44:01]

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
File tools/json_schema_compiler/idl_schema.py (right):

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:208: 'with a non-void return type')
On 2015/04/01 22:46:19, Devlin wrote:
> including "name" here would probably be helpful if this fails.

Done.

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:234: properties['enum'] =
[self.cast_from_json_type(properties['type'], enum)
On 2015/04/01 22:46:19, Devlin wrote:
> why not:
> map(lambda e: self.cast_from_json_type(properties['type'], e), enum_values)?

List comprehensions are preferred over map for anything non-trivial:
https://google-styleguide.googlecode.com/svn/trunk/pyguide.html#Deprecated_La...

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:411: # it as a list. Convert back to
key-value pairs.
On 2015/04/01 22:46:19, Devlin wrote:
> Nit: put this comment two lines lower (above where you do the conversion).

you could argue it either way (the reason why it's called "properties_as_list"
is because the return type is a list) but sure.

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:414: for prop in properties_as_list:
On 2015/04/01 22:46:19, Devlin wrote:
> Would this be crazy?
> map(lambda p: self.properties[p.pop('name')] = p, properties_as_list)

Yes, that would be crazy, however - good point, I can inline the
prop.pop('name').

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:416: self.properties[name] = prop
On 2015/04/01 23:12:54, Red Daly wrote:
> Does the IDL parser ensure there are no duplicate names?

Good question - and I just tested, and it doesn't. The effect here would be to
only take the last declaration. We could add validation, which might help in
rare cases, though we don't validate other places this would also be a problem
(e.g. duplicate function names - less of an issue because they're tuples in a
list, I suppose).

[Devlin, 2015-04-01 23:53:12]

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
File tools/json_schema_compiler/idl_schema.py (right):

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:416: self.properties[name] = prop
On 2015/04/01 23:44:00, kalman wrote:
> On 2015/04/01 23:12:54, Red Daly wrote:
> > Does the IDL parser ensure there are no duplicate names?
> 
> Good question - and I just tested, and it doesn't. The effect here would be to
> only take the last declaration. We could add validation, which might help in
> rare cases, though we don't validate other places this would also be a problem
> (e.g. duplicate function names - less of an issue because they're tuples in a
> list, I suppose).

Enh, why not assert that we don't.  Doesn't hurt.

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
File tools/json_schema_compiler/test/idl_properties.idl (right):

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/test/idl_properties.idl:17: // Unvalued property.
On 2015/04/01 22:46:20, Devlin wrote:
> Why do we want to allow this?

Nudge.

[not at google - send to devlin, 2015-04-02 00:10:33]

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
File tools/json_schema_compiler/idl_schema.py (right):

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/idl_schema.py:416: self.properties[name] = prop
On 2015/04/01 23:53:12, Devlin wrote:
> On 2015/04/01 23:44:00, kalman wrote:
> > On 2015/04/01 23:12:54, Red Daly wrote:
> > > Does the IDL parser ensure there are no duplicate names?
> > 
> > Good question - and I just tested, and it doesn't. The effect here would be
to
> > only take the last declaration. We could add validation, which might help in
> > rare cases, though we don't validate other places this would also be a
problem
> > (e.g. duplicate function names - less of an issue because they're tuples in
a
> > list, I suppose).
> 
> Enh, why not assert that we don't.  Doesn't hurt.

Done.

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
File tools/json_schema_compiler/test/idl_properties.idl (right):

https://codereview.chromium.org/1051563002/diff/40001/tools/json_schema_compi...
tools/json_schema_compiler/test/idl_properties.idl:17: // Unvalued property.
On 2015/04/01 23:53:12, Devlin wrote:
> On 2015/04/01 22:46:20, Devlin wrote:
> > Why do we want to allow this?
> 
> Nudge.

You're still allowed properties without values (e.g. chrome.runtime.id).

[Devlin, 2015-04-02 00:16:29]

lgtm

[not at google - send to devlin, 2015-04-02 00:21:17]

The CQ bit was checked by kalman@chromium.org

[commit-bot: I haz the power, 2015-04-02 00:23:35]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1051563002/80001

[commit-bot: I haz the power, 2015-04-02 02:04:16]

Committed patchset #5 (id:80001)

[commit-bot: I haz the power, 2015-04-03 20:21:23]

Patchset 5 (id:??) landed as
https://crrev.com/d94f954f5e1c32e7be85e2ba835a143fd388a0d8
Cr-Commit-Position: refs/heads/master@{#323401}