Pvl

------------------

Fixed
+++++
* The parser was requesting the next token after an end-statement, even
though nothing was done with this token (in the future it could
be a comment that should be processed). In the very rare case
where all of the "data" bytes in a file with an attached PVL label
(like a .IMG or .cub file) actually convert to UTF with no
whitespace characters, that next token will take an unacceptable
amount of time to return, if it does at all. The parser now does
not request additional tokens once an end-statement is identified
(Issue 104).

------------------

Fixed
+++++
* Deeply nested Aggregation Blocks (Object or Group) which had mis-matched
Block Names should now properly result in LexerErrors instead of
resulting in StopIteration Exceptions (Issue 100).

* The default "Omni" parsing strategy, now considers the ASCII NULL character
("\0") a "reserved character." The practical effect is that the
ASCII NULL can not be in parameter names or unquoted strings (but
would still be successfully parsed in quoted strings). This means
that PVL-text that might have incorrectly used ASCII NULLs as
delimiters will once again be consumed by our omnivorous parser
(Issue 98).

------------------

Added
+++++
* pvl.collections.Quantity objects now have __int__() and __float__()
functions that will return the int and float versions of their
.value parameter to facilitate numeric operations with Quantity
objects (Issue 91).
* pvl.load() now has an `encoding=` parameter that is identical in usage
to the parameter passed to `open()`, and will attempt to decode the whole
file as if it had been encoded thusly. If it encounters a decoding error,
it will fall back to decoding the bytes one at a time as ASCII text (Issue 93).

Fixed
+++++
* If the PVL-text contained characters beyond the set allowed by the
PVL specification, the OmniGrammar would refuse to parse them.
This has been fixed to allow any valid character to be parsed,
so that if there are weird UTF characters in the PVL-text, you'll get
those weird UTF characters in the returned dict-like. When the
stricter PVL, ODL, or PDS3 dialects are used to "load" PVL-text,
they will properly fail to parse this text (Issue 93).
* Empty parameters inside groups or objects (but not at the end), would
cause the default "Omni" parsing strategy to go into an infinite
loop. Empty parameters in PVL, ODL, and PDS3 continue to not be
allowed (Issue 95).

------------------

Added
+++++
* So many tests, increased coverage by about 10%.

Fixed
+++++
* Attempting to import `pvl.new` without *multidict* being available,
will now properly yield an ImportError.
* The `dump()` and `dumps()` functions now properly overwritten in `pvl.new`.
* All encoders that descended from PVLEncoder didn't properly have group_class and
object_class arguments to their constructors, now they do.
* The `char_allowed()` function in grammar objects now raises a more useful ValueError
than just a generic Exception.
* The new `collections.PVLMultiDict` wasn't correctly inserting Mapping objects with
the `insert_before()` and `insert_after()` methods.
* The `token.Token` class's `__index__()` function didn't always properly return an
index.
* The `token.Token` class's `__float__()` function would return int objects if the
token could be converted to int. Now always returns floats.

------------------

Added
+++++
* Added a default_timezone parameter to grammar objects so that they could
both communicate whether they had a default timezone (if not None),
and what it was.
* Added a pvl.grammar.PDSGrammar class that specifies the default UTC
time offset.
* Added a pvl.decoder.PDSLabelDecoder class that properly enforces only
milisecond time precision (not microsecond as ODL allows), and does
not allow times with a +HH:MM timezone specifier. It does assume
any time without a timezone specifier is a UTC time.
* Added a ``real_cls`` parameter to the decoder classes, so that users can specify
an arbitrary type with which real numbers in the PVL-text could be returned in
the dict-like from the loaders (defaults to ``float`` as you'd expect).
* The encoders now support a broader range of real types to complement the decoders.

Changed
+++++++
* Improved some build and test functionality.
* Moved the is_identifier() static function from the ODLEncoder to the ODLDecoder
where it probably should have always been.


Fixed
+++++
* Very long Python ``str`` objects that otherwise qualified as ODL/PDS3 Symbol Strings,
would get written out with single-quotes, but they would then be split across lines
via the formatter, so they should be written as Text Strings with double-quotes.
Better protections have been put in place.
* pvl.decoder.ODLDecoder now will return both "aware" and "naive"
datetime objects (as appropriate) since "local" times without a
timezone are allowed under ODL.
* pvl.decoder.ODLDecoder will now properly reject any unquoted string
that does not parse as an ODL Identifier.
* pvl.decoder.ODLDecoder will raise an exception if there is a seconds value
of 60 (which the PVLDecoder allows)
* pvl.encoder.ODLEncoder will raise an exception if given a "naive" time
object.
* pvl.encoder.PDSLabelEncoder will now properly raise an exception if
a time or datetime object cannot be represented with only milisecond
precision.

------------------

Added
+++++
* Modified `pvl_validate` to more robustly deal with errors, and also provide
more error-reporting via `-v` and `-vv`.
* Modified ISISGrammar so that it can parse comments that begin with an octothorpe ().

Fixed
+++++
* Altered documentation in grammar.py that was incorrectly indicating that
there were parameters that could be passed on object initiation that would
alter how those objects behaved.