Format.CQL for seralizing and deserializing filters

[tschaub]

To facilitate the creation of OpenLayers.Filter instances, a parser could be created that reads CQL (OGC Common Query Language). The same parser should write CQL given a filter.

CQL is defined as part of the OGC Catalog specification. Some references below:

•  OGC Catalog Specification
•  uDig docs for CQL
•  GeoTools docs for CQL

[dwinslow]

patch for CQL parsing

[dwinslow]

I've attached a patch which provides parsing/encoding for a subset of CQL syntax.

[erilem]

Replying to dwinslow:

I've attached a patch which provides parsing/encoding for a subset of CQL syntax.

Missing files in the patch (I hope).

[dwinslow]

Replying to erilem:

Replying to dwinslow:

I've attached a patch which provides parsing/encoding for a subset of CQL syntax.

Missing files in the patch (I hope).

Oops, sorry about that. Attaching a better patch shortly.

[dwinslow]

patch including missing files from original

[tschaub]

Reviewing now.

[tschaub]

This will make a really great addition. In reviewing, I've caught a few things and added tests. There is still one failing test. Here's what I've done so far:

• removed trailing commas to get it to load in IE
• added semicolons where appropriate (improves things with our limited compressor)
• updated example to allow users to change CQL and see results
• added CQL.js format to OpenLayers.js loader
• added docs
• made it so comparison filters are written with quoted strings only (numbers were quoted as well)
• added write support for WITHIN, INTERSECTS, CONTAINS
• replaced use of bracket notation with charAt for strings (the former doesn't work in IE and was causing infinite recursion)
• made infinite recursion less likely in the balanced parentheses matcher for geometries
• made it possible to use spatial filters in a logical filter - this required adding SPATIAL before PROPERTY to the list of follows for LOGICAL (tests pass, but this needs confirmation as the right thing to do)
• made it so spatial filters store geometries instead of features

[tschaub]

In adding write tests, I've gotten it so everything passes except for the writing the filter that results from "A = '1' AND (B = '2' OR C = '3')". Seems to me like we could go for the easy solution and wrap all subfilters or keep track of precedence within logical operators.

[tschaub]

read and write cql

[tschaub]

Ok, latest patch avoids the logical operator precedence issue by wrapping all subfilters of a logical filter in parens when writing. Tests cover this.

[tschaub]

Ok, with tests passing and examples working in IE 6, FF 3, and Chrome 9, I think this is good to go. This dramatically simplifies filter creation and I hope it gets heavy use. It will also make a nice alternative filterToParams method for the HTTP protocol (see #3032).

Thanks dwins for the great contribution.

[tschaub]

In w/ r11065.