Skip to main content

Python implementation for PVL (Parameter Value Language) parsing and encoding.

Project description


Documentation Status Codecov coverage PyPI version PyPI Downloads/month conda-forge version conda-forge downloads

Python implementation of a PVL (Parameter Value Language) library.

PVL is a markup language, like JSON or YAML, commonly employed for entries in the Planetary Data System used by NASA to archive mission data, among other uses. This package supports both encoding and decoding a variety of PVL ‘flavors’ including PVL itself, ODL, NASA PDS 3 Labels, and USGS ISIS Cube Labels.


Can either install with pip or with conda.

To install with pip, at the command line:

$ pip install pvl

Directions for installing with conda-forge:

Installing pvl from the conda-forge channel can be achieved by adding conda-forge to your channels with:

conda config --add channels conda-forge

Once the conda-forge channel has been enabled, pvl can be installed with:

conda install pvl

It is possible to list all of the versions of pvl available on your platform with:

conda search pvl --channel conda-forge

Basic Usage

pvl exposes an API familiar to users of the standard library json module.

Decoding is primarily done through pvl.load() for file-like objects and pvl.loads() for strings:

>>> import pvl
>>> module = pvl.loads("""
...     foo = bar
...     items = (1, 2, 3)
...     END
... """)
>>> print(module)
  ('foo', 'bar')
  ('items', [1, 2, 3])
>>> print(module['foo'])

There is also a pvl.loadu() to which you can provide the URL of a file that you would normally provide to pvl.load().

You may also use pvl.load() to read PVL text directly from an image that begins with PVL text:

>>> import pvl
>>> label = pvl.load('tests/data/pattern.cub')
>>> print(label)
   {'Core': {'Dimensions': {'Bands': 1,
                            'Lines': 90,
                            'Samples': 90},
             'Format': 'Tile',
             'Pixels': {'Base': 0.0,
                        'ByteOrder': 'Lsb',
                        'Multiplier': 1.0,
                        'Type': 'Real'},
             'StartByte': 65537,
             'TileLines': 128,
             'TileSamples': 128}})
  ('Label', PVLObject([
    ('Bytes', 65536)
>>> print(label['IsisCube']['Core']['StartByte'])

Similarly, encoding Python objects as PVL text is done through pvl.dump() and pvl.dumps():

>>> import pvl
>>> print(pvl.dumps({
...     'foo': 'bar',
...     'items': [1, 2, 3]
... }))
FOO   = bar
ITEMS = (1, 2, 3)

pvl.PVLModule objects may also be pragmatically built up to control the order of parameters as well as duplicate keys:

>>> import pvl
>>> module = pvl.PVLModule({'foo': 'bar'})
>>> module.append('items', [1, 2, 3])
>>> print(pvl.dumps(module))
FOO   = bar
ITEMS = (1, 2, 3)

A pvl.PVLModule is a dict-like container that preserves ordering as well as allows multiple values for the same key. It provides similar semantics to a list of key/value tuples but with dict-style access:

>>> import pvl
>>> module = pvl.PVLModule([
...     ('foo', 'bar'),
...     ('items', [1, 2, 3]),
...     ('foo', 'remember me?'),
... ])
>>> print(module['foo'])
>>> print(module.getall('foo'))
['bar', 'remember me?']
>>> print(module.items())
  ('foo', 'bar')
  ('items', [1, 2, 3])
  ('foo', 'remember me?')
>>> print(pvl.dumps(module))
FOO   = bar
ITEMS = (1, 2, 3)
FOO   = 'remember me?'

However, there are some aspects to the default pvl.PVLModule that are not entirely aligned with the modern Python 3 expectations of a Mapping object. If you would like to experiment with a more Python-3-ic object, you could instantiate a pvl.collections.PVLMultiDict object, or import as pvl in your code to have the loaders return objects of this type (and then easily switch back by just changing the import statement). To learn more about how PVLMultiDict is different from the existing OrderedMultiDict that PVLModule is derived from, please read the new PVLMultiDict documentation.

The intent is for the loaders (pvl.load(), pvl.loads(), and pvl.loadu()) to be permissive, and attempt to parse as wide a variety of PVL text as possible, including some kinds of ‘broken’ PVL text.

On the flip side, when dumping a Python object to PVL text (via pvl.dumps() and pvl.dump()), the library will default to writing PDS3-Standards-compliant PVL text, which in some ways is the most restrictive, but the most likely version of PVL text that you need if you’re writing it out (this is different from pre-1.0 versions of pvl).

You can change this behavior by giving different parameters to the loaders and dumpers that define the grammar of the PVL text that you’re interested in, as well as custom parsers, decoders, and encoders.

For more information on custom serilization and deseralization see the full documentation.


Feedback, issues, and contributions are always gratefully welcomed. See the contributing guide for details on how to help and setup a development environment.


All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

When updating this file, please add an entry for your change under Not Yet Released and one of the following headings:

  • Added - for new features.

  • Changed - for changes in existing functionality.

  • Deprecated - for soon-to-be removed features.

  • Removed - for now removed features.

  • Fixed - for any bug fixes.

  • Security - in case of vulnerabilities.

If the heading does not yet exist under Not Yet Released, then add it as a 3rd level heading, underlined with pluses (see examples below).

When preparing for a public release add a new 2nd level heading, underlined with dashes under Not Yet Released with the version number and the release date, in year-month-day format (see examples below).

Not Yet Released

1.3.2 (2022-02-05)


  • 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).

1.3.1 (2022-02-05)


  • 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).

1.3.0 (2021-09-10)


  • 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).


  • 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).

1.2.1 (2021-05-31)


  • So many tests, increased coverage by about 10%.


  • Attempting to import without multidict being available, will now properly yield an ImportError.

  • The dump() and dumps() functions now properly overwritten in

  • 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.

1.2.0 (2021-03-27)


  • 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.


  • 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.


  • 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.

1.1.0 (2020-12-04)


  • 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 (#).


  • Altered documentation in that was incorrectly indicating that there were parameters that could be passed on object initiation that would alter how those objects behaved.

1.0.1 (2020-09-21)


  • The PDSLabelEncoder was improperly raising an exception if the Python datetime object to encode had a tzinfo component that had zero offset from UTC.

1.0.0 (2020-08-23)

This production version of the pvl library consists of significant API and functionality changes from the 0.x version that has been in use for 5 years (a credit to Trevor Olson’s skills). The documentation has been significantly upgraded, and various granular changes over the 10 alpha versions of 1.0.0 over the last 8 months are detailed in their entries below. However, here is a high-level overview of what changed from the 0.x version:


  • pvl.load() and pvl.dump() take all of the arguments that they could take before (string containing a filename, byte streams, etc.), but now also accept any os.PathLike object, or even an already-opened file object.

  • pvl.loadu() function will load PVL text from URLs.

  • Utility programs pvl_validate and pvl_translate were added, please see the “Utility Programs” section of the documentation for more information.

  • The library can now parse and encode PVL Values with Units expressions with third-party quantity objects like astropy.units.Quantity and pint.Quantity. Please see the “Quantities: Values and Units” section of the documentation.

  • Implemented a new PVLMultiDict (optional, needs 3rd party multidict library) which which has more pythonic behaviors than the existing OrderedMultiDict. Experiment with getting it returned by the loaders by altering your import statement to import as pvl and then using the loaders as usual to get the new object returned to you.


  • Only guaranteed to work with Python 3.6 and above.

  • Rigorously implemented the three dialects of PVL text: PVL itself, ODL, and the PDS3 Label Standard. There is a fourth de-facto dialect, that of ISIS cube labels that is also handled. Please see the “Standards & Specifications” section of the documentation.

  • There is now a default dialect for the dump functions: the PDS3 Label Standard. This is different and more strict than before, but other output dialects are possible. Please see the “Writing out PVL text” section in the documentation for more information, and how to enable an output similar to the 0.x output.

  • There are now pvl.collections and pvl.exceptions modules. There was previously an internal pvl._collections module, and the exception classes were scattered through the other modules.


  • All datetime.time and datetime.datetime objects returned from the loaders are now timezone “aware.” Previously some were and some were not.

  • Functionality to correctly parse dash (-) continuation lines in ISIS output is now supported.

  • The library now properly parses quoted strings that include backslashes.


  • The pvl.collections.Units object is deprecated in favor of the new pvl.collections.Quantity object (really a name-only change, no functionality difference).

1.0.0-alpha.9 (2020-08-18)

  • Minor addition to pvl.collections.MutableMappingSequence.

  • Implemented PVLMultiDict which is based on the 3rd Party multidict.MultiDict object as an option to use instead of the default OrderedMultiDict. The new PVLMultiDict is better aligned with the Python 3 way that Mapping objects behave.

  • Enhanced the existing OrderedMultiDict with some functionality that extends its behavior closer to the Python 3 ideal, and inserted warnings about how the retained non-Python-3 behaviors might be removed at the next major patch.

  • Implemented that can be included for those that wish to try out what getting the new PVLMultiDict returned from the loaders might be like by just changing an import statement.

1.0.0-alpha.8 (2020-08-01)

  • Renamed the _collections module to just collections.

  • Renamed the Units class to Quantity (Units remains, but has a deprecation warning).

  • Defined a new ABC: pvl.collections.MutableMappingSequence

  • More detail for these changes can be found in Issue #62.

1.0.0-alpha.7 (2020-07-29)

  • Created a new module and grouped all pvl Exceptions there. Addresses #58

  • Altered the message that LexerError emits to provide context around the character that caused the error.

  • Added bump2version configuration file.

1.0.0-alpha.6 (2020-07-27)

  • Enforced that all datetime.time and datetime.datetime objects returned should be timezone “aware.” This breaks 0.x functionality where some were and some weren’t. Addresses #57.

1.0.0-alpha.5 (2020-05-30)

  • ISIS creates PVL text with unquoted plus signs (“+”), needed to adjust the ISISGrammar and OmniGrammar objects to parse this properly (#59).

  • In the process of doing so, realized that we have some classes that optionally take a grammar and a decoder, and if they aren’t given, to default. However, a decoder has a grammar object, so if a grammar isn’t provided, but a decoder is, the grammar should be taken from the decoder, otherwise you could get confusing behavior.

  • Updated pvl_validate to be explicit about these arguments.

  • Added a –version argument to both pvl_translate and pvl_validate.

1.0.0.-alpha.4 (2020-05-29)

  • Added the pvl.loadu() function as a convenience function to load PVL text from URLs.

1.0.0-alpha.3 (2020-05-28)

  • Implemented tests in tox and Travis for Python 3.8, and discovered a bug that we fixed (#54).

1.0.0-alpha.2 (2020-04-18)

  • The ability to deal with 3rd-party ‘quantity’ objects like astropy.units.Quantity and pint.Quantity was added and documented, addresses #22.

1.0.0-alpha.1 (2020-04-17)

This is a bugfix on 1.0.0-alpha to properly parse scientific notation and deal with properly catching an error.

1.0.0-alpha (winter 2019-2020)

This is the alpha version of release 1.0.0 for pvl, and the items here and in other ‘alpha’ entries may be consolidated when 1.0.0 is released. This work is categorized as 1.0.0-alpha because backwards-incompatible changes are being introduced to the codebase.

  • Refactored code so that it will no longer support Python 2, and is only guaranteed to work with Python 3.6 and above.

  • Rigorously implemented the three dialects of PVL text: PVL itself, ODL, and the PDS3 Label Standard. There is a fourth de-facto dialect, that of ISIS cube labels that is also handled. These dialects each have their own grammars, parsers, decoders, and encoders, and there are also some ‘Omni’ versions of same that handle the widest possible range of PVL text.

  • When parsing via the loaders, pvl continues to consume as wide a variety of PVL text as is reasonably possible, just like always. However, now when encoding via the dumpers, pvl will default to writing out PDS3 Label Standard format PVL text, one of the strictest dialects, but other options are available. This behavior is different from the pre-1.0 version, which wrote out more generic PVL text.

  • Removed the dependency on the six library that provided Python 2 compatibility.

  • Removed the dependency on the pytz library that provided ‘timezone’ support, as that functionality is replaced with the Standard Library’s datetime module.

  • The private pvl/ file was removed, as its capability is now accomplished with the Python Standard Library.

  • The private pvl/ file was removed, as its capability is now accomplished with the Standard Library’s datetime module.

  • the private pvl/ file was removed, as its capabilities are now mostly replaced with the new grammar module and some functions in other new modules.

  • Internally, the library is now working with string objects, not byte literals, so the pvl/ module is no longer needed.

  • Added an optional dependency on the 3rd party dateutil library, to parse more exotic date and time formats. If this library is not present, the pvl library will gracefully fall back to not parsing more exotic formats.

  • Implmented a more formal approach to parsing PVL text: The properties of the PVL language are represented by a grammar object. A string is broken into tokens by the lexer function. Those tokens are parsed by a parser object, and when a token needs to be converted to a Python object, a decoder object does that job. When a Python object must be converted to PVL text, an encoder object does that job.

  • Since the tests in tests/ and tests/ were really just exercising the loader and dumper functions, those tests were moved to tests/, but all still work (with light modifications for the new defaults). Unit tests were added for most of the new classes and functions. All docstring tests now also pass doctest testing and are now included in the make test target.

  • Functionality to correctly parse dash (-) continuation lines written by ISIS as detailed in #34 is implemented and tested.

  • Functionality to use pathlib.Path objects for pvl.load() and pvl.dump() as requested in #20 and #31 is implemented and tested.

  • Functionality to accept already-opened file objects that were opened in ‘r’ mode or ‘rb’ mode as alluded to in #6 is implemented and tested.

  • The library now properly parses quoted strings that include backslashes as detailed in #33.

  • Utility programs pvl_validate and pvl_translate were added.

  • Documentation was updated and expanded.

0.3.0 (2017-06-28)

  • Create methods to add items to the label

  • Give user option to allow the parser to succeed in parsing broken labels

0.2.0 (2015-08-13)

  • Drastically increase test coverage.

  • Lots of bug fixes.

  • Add Cube and PDS encoders.

  • Cleanup README.

  • Use pvl specification terminology.

  • Added element access by index and slice.

0.1.1 (2015-06-01)

  • Fixed issue with reading Pancam PDS Products.

0.1.0 (2015-05-30)

  • First release on PyPI.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pvl-1.3.2.tar.gz (153.1 kB view hashes)

Uploaded Source

Built Distribution

pvl-1.3.2-py2.py3-none-any.whl (66.1 kB view hashes)

Uploaded Python 2 Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page