Fix some of the fixmes in _base/xhr.js, and re-org to allow other io transports to use xhr infrastructure

[James Burke]

Asking Alex to review:

The patch for this ticket fixes some of the fixmes in _base/xhr.js, in particular:

• Preserves the user's input args
• Better timeout support, with a "dojoType" property added to the Error object.
• Make sure we xhr abort on an deferred error callback.

This patch also makes two methods publicly visible so they can be used by other io transports: dojo._ioWatch (was private _launch) and dojo._ioSetArgs (was part of private _setupForXhr).

[James Burke]

Patch file #1

[James Burke]

File sizes of rhino-compressed dojo.js:

before patch: 64238
after patch:  64849

This patch adds 611 bytes.

[James Burke]

Just added 2924-2.patch. This version has support for passing load and error callbacks as part of the args object passed to the public APIs. The load and error callbacks are added to the Deferred object underneath, and when they are called, they have their "this" set to the args object. They also receive the ioArgs as a second argument (the deferred result is the first).

This should make it very easy to make encapsulated IO calls that do not want to hold onto their args objects just so they can add callbacks to the Deferred object. Also, it allows easy access to the args state, since it is the "this" value in those callbacks.

File sizes of rhino-compressed dojo.js:

before patch:  64238
after patch2:  65099

Patch 2 adds 861 bytes compared to the unpatched version. It is 150 bytes larger than Patch 1.

I think that 150 bytes is worth avoiding having to explain Deferreds for the simple use cases (and not to mention the extra bytes the user will have to type to use the Deferreds directly).

This change would also work for the dojo.rpc case (the load and error callbacks are only added to the Deferred if they are functions).

[James Burke]

New version with load/error support, and puts form parsing inside dojo._ioSetArgs

[James Burke]

Just updated Patch 2 -- when trying to use the xhr infrasctructure for dojo.io.script (see ticket #2911), it seemed best to put the form parsing inside dojo._ioSetArgs().

File sizes of rhino-compressed dojo.js:

before patch:  64238
after patch2:  65016

Patch 2 adds 778 bytes compared to the unpatched version. It is 167 bytes larger than Patch 1. (Previous comment said the original Patch 2 was 150 bytes bigger than Patch 1, but that comment should have read "250 bytes bigger").

[alex]

(In [8488]) merging the first part James' patch to add ScriptSrc? IO to Core. Refs #2924

[alex]

thanks for getting this revved and done. Marking fixed.

[alex]

(In [8493]) fixes the tests. Refs #2924

[alex]

(In [8540]) fixing XHR Post regression added in last rev. Refs #2924

[(none)]

Milestone 0.9M2 deleted

[Adam Peller]

(In [16566]) Change console.debug call to console.error, refs #2924, fixes #8583. (Do we really need a console message here at all if the error can be detected programatically? Style change refs #7250