* Added error output to test_program.go
* Added multi-line literal string support to lexer
* Added multi-line string supprt to lexer
* Added unit-test for new string support
* Modified test.sh to take an optional parameter to run an individual BurntSushi test suite.
* Fixed formatting
* Refactored type names and file names to mesh with existing TOML library more closely
* Added QueryResult structure that provides values and position data
* Added Query() method to TomlTree type
* Tests, tests, and more tests
* Fixed bug where positions returned from some tables were invalid
* Added test case for bug patch
The bugfix was an interesting case. Position information wasn't being
set in cases where createPath was called. So table names like [foo.bar]
would result in table 'foo' having no position.
QueryResult now stores result items and position data, which aligns more
strongly with the rest of the library features than a plain
[]interface[}. The design of the parser_test unittest was revised to
use array/map/scalar serialization (like match_test), since Go 1.3
redesigned maps to randomly order their keys. Since naive comparisons of
map data is now no longer possible, the unittest now sorts map
keys:value combinations.
* Patched a bug where getPosition("") was returning an invalid Position
* Revised parser_test to use serialization for comparisons for Go 1.3
* Added public Query interface
* Added filter function callback support
* Added "script" function callback support
Queries are generated via Compile(), which then may be run via Execute()
as many times as needed. Much like compiling a regex, this is done to
elide the need to re-parse and build the funciton tree for each
execution.
The distinction between 'filter' and 'script' is borrowed from their
syntactic equivalents in jsonpath. Right now, these accept no arguments
in the query, and instead merely pass the current node to the callback.
Filters return a bool and determine if the node is kept or culled out.
'Scripts' return a string or an in64, which is in turn used in an index
or key filter (respectively) on the current node's data.
A few callbacks are provided by default, with the ability to add
additional callbacks before calling Execute() on a compiled query.
As it turns out, closures are very hard to validate without running them.
Since table-driven tests tend to rely on value types that can be
compared directly, using structs that adhere to a generic callback
interface is more work, but is more easily tested.
* Changed jsonpath match functions to structs with Call() methods
* Added tests to verify the generation of jsonpath QueryPath data
* Added tests to verify jsonpath lexer
* Fixed jsonpath whitespace handling bug
* Fixed numerous flaws in jsonpath parser
TomlDocument provides an optional TOML processing path where position
informaiton is stored alongside a TomlTree.
* Added Position struct
* Revised TomlTree to contain position data
* Added tomlValue to bind positions to values
* Revised parser to emit position data
* Revised token to use new Position struct
* Added tests for new functionality
* Bugfixed table array duplicate key handling
* Applied gofmt to all code
Thomas Pelletier