pykwalify

New keywords:

• Added support for keyword nullable. It is now possible to allow for a key not to be empty, when required keyword is not used.
• Added support for keyword class. It will not cause any validation errors, but serves to make kwalify schemas compatible that uses that keywork.

Bug fixes:

• Improved compatibility with unicodes to validate as strings.

Changed behaviour:

• Propergate json and yaml loading errors when used from the cli to the user for easier debugging.

General changes:

• Allow ruamel.yaml versions up to 0.16
• License is now bundled with the built release.

New keywords:

• Add support for keyword example. It does nothing and have no validation done on it.
• Add support for keyword version. It does nothing and have no validation done on it.
• Add support for keyword date and added support keyword format. This can be used to validate many different types of datetime objects.
• Add support for keyword length. It is very similar to range but works primarily string types.
• Add support for keyword assert. It works by running the python code assert and check if any exception is raised.
  This feature is considered dangerouns becuase there is only simple logic to prevent escaping out from validation.

Bug fixes:

• Fixed a bug where regexes marked as 'required' in a map were matched as strings, rather than regexes.
• Fixed a bug where the type validation did not work when schema specefied a sequence of map objects. It now outputs a proper ...is not a dict... error instead.
• Fixed a bug in unique validation when a key that it tried to lookup in the data would not exists.
  Now it just ignores that it did not find any value becuase a missing value do not impact validation.
• Fixed a bug with keyword ident when the rule value was verified to be a boolean. It now only accepts boolean values as expected.
• Fixed a bug where if allowempty was specefied in a mapping type inside a sequence type then it would not properly validate.
• Fixed a bug where loaded extensions would not allways work in complex & nested objects.
• Fixed a major bug in very nested seq schemas where if the schema expected another seq but the value was something else it would not raise it as a validation error.
  This has now been fixed and now raises the proper error.
• Fixed a bug where include directive would not work in all cases when used inside a key in a mapping block.

New features:

• It is now possible to specify a default rule when using a mapping.
  The rule will be used whenever no other key could be found.
  This is a port of a missing feature from original kwalify implementation.
• Added new helper method keywords to Rule class that can output all active keywords for any Rule object.
  This helps when debugging code to be able to easily dump what all active keywords for any Rule object.
• Added new cli flag --strict-rule-validation that will validate that all used keywords in all Rule objects only uses the rules that is supported by each defined type.
  If you only use a Core object then set strict_rule_validation=True when creating the Core object instance.
  This feature is opt-in in this releaes but will be mandatory in releases >= 1.7.0.
• Added new cli flag --fix-ruby-style-regex that will trim slashes from ruby style regex/patterns.
  When using this flag the first and last / will be trimmed of the pattern before running validation.
  If you only use a Core object then set fix_ruby_style_regex=True when creating the Core object instance.
  Default behaviour will still be that you should use python style regex values but this flag can help in some cases when you can't change the schema.
• Added new cli flag --allow-assertions that will enable the otherwise blocked keyword assert.
  This flag was introduced so that pykwalify would not assert assertions without user controll.
  Default behaviour will be to raise a CoreError is assertion is used but not allowed explicitly.
  If you only use a Core object then set allow_assertions=True when creating the Core object instance.

Changed behaviour:

• Removed the force of UTF-8 encoding when importing pykwalify package. It caused issues with jypiter notebooks on python 2.7.x
  Added documentation in Readme regarding the suggested solution to use PYTHONIOENCODING=UTF-8 if the default solution do not work.
• Validation do no longer continue to process things like pattern, regex, timestamp, range and other additional checks
  if the type check fails. This can cause problems where previous errors will now initially be silenced when the typecheck for
  that value fails, but reappear again when the type check is fixed. (sbrunner)
• Catches TypeError when doing regex validation. That happens when the value is not a parsable string type.
• Checking that the value is a valid dict object is now done even if the mapping keyword is not specefied in the schema.
  This makes that check more eager and errors can apear that previously was not there.
• Changed the sane default type if that key is not defined to be str. Before this, type had to be defined every time and the default type did not work as expected.
  This is a major change and can cause validation to either fail or to stop failing depending on the case.
• Changed validation for if a value is required and a value in a list for example is None. It now adds a normal validation errors instead of raising a CoreError.
• Value for keyword desc now MUST be a string or a RuleError will be raised.
• Value for keyword example now MUST be a string or a RuleError will be raised.
• Value for keyword name now MUST be a string or a RuleError will be raised.

General changes:

• Ported alot of testcases directly from Kwalify test data (test-validator.yaml -> 30f.yaml & 43s.yaml) so that this lib can have greater confidence that rules is implemented in the same way as Kwalify.
• Refactored test_core_files method to now accept test files with multiple of documents. The method now tries to read all documents from each test file and run each document seperatly.
  It now alos reports more detailed about what file and document that fails the test to make it easier to track down problems.
• Major refactoring of test files to now be grouped based on what they are testing instead of a increased counter that do not represent anything.
  It will be easier to find out what keywords lack tests and what keywords that have enough tests.

• Convert all documentation to readthedocs
• True/False is no longer considered valid integer
• python3 'bytes' objects is now a valid strings and text type
• The regular PyYaml support is now deprecated in favor of ruamel.yaml, see the following link for more details about
  PyYaml being deprecated https://bitbucket.org/xi/pyyaml/issues/59/has-this-project-been-abandoned
  PyYaml will still be possible to use in the next major release version (1.6.0) but removed in release (1.7.0) and forward.
• ruamel.yaml is now possible to install with the following command for local development pip install -e '.[ruamel]'
  and for production, use pip install 'pykwalify[ruamel]'
• ruamel.yaml is now used before PyYaml if installed on your system
• Fixed a bug where scalar type was not validated correctly.
• Unpin all dependencies but still maintain a minimum versions of each lib
• Allowed mixing of regex and normal keywords when matching a string (jmacarthur)

• Improvements to documentation (scottclowe).
• Improved code linting by reworking private variables in Rule class to now be properties and updated
  all code that used the old way.
• Improved code linting by reworking all Log messages to render according to pep standard.
  (By using %s and passing in variables as positional arguments)
• Fix bug when validating sequence and value should only be unicode escaped when a string
• Improve validation of timestamps.
• Improve float validation to now accept strings that is valid ints that uses scientific notation, "1e-06" for example.
• Update travis to test against python 3.6

• float / number type now support range restrictions
• ranges on non number types (e.g. seq, string) now need to be non negative.
• Fixed encoding bug triggered when both regex matching-rule 'any' and 'all' found keyword that
  failed regex match. Added failure unit tests to cover regex matching-rule 'any' and 'all' during
  failed regex match. Updated allowed rule list to include matching-rule 'all'.
• Changed _validate_mappings method from using re.match to re.search. This fixes bug related to
  multiple keyword regex using matching-rule 'any'. Added success unit tests to test default, 'any',
  and 'all' matching-rule.

• Added tests to sdist to enable downstream packaging to run tests. No code changes in this release.

• Dropped support for python 3.2 becuase of unicode literals do not exists in python 3.2.
• Fixed logging & raised exceptions when using unicode characters inside schemas/data/filenames.
• Reworked all RuleError exceptions to now have better exception messages.
• RuleError exceptions now have a unique 'error_key' that can make it easier to identify what error it is.
• Paths for RuleErrors have been moved inside the exception as a variable.
• Rewrote all SchemaConflict exceptions to be more human readable.

• Rewrote most of the error messages to be more human readable. See docs/Upgrade Instructions.md
  for more details.
• It is now possible to use the exceptions that was raised for each validation error. It can be
  found in the variable c.validation_errors_exceptions. They contain more detailed information
  about the error.

• This feature is NEW and EXPERIMENTAL. Implemented support for multiple values inside in a sequence. This will allow the defenition of different types that one sequence can contain. You can either require each value in the sequence to be valid against one to all of the different possibilities. Tests show that it still maintains backward compatibility with all old schemas but it can't be guarantee. If you find a regression in this release please file a bug report so it can be fixed ASAP.

• This feature is NEW and EXPERIMENTAL. Added ability to define python files that can be used to have custom python code/functions that can be called on all types so that custom/extra validation can be done on all data structures.

• Add new keyword 'func' that is a string and is used to point to a function loaded via the extension system.

• Add new keyword 'extensions' that can only be used on the top level of the schema. It is should be a list with strings of files that should be loaded by the extension system. Paths can be relative or absolute.

• New cli option '-e FILE' or '--extension FILE' that can be used to load extension files from cli.

• Fixed a bug where types did not raise exceptions properly. If schema said it should be a map but data was a sequence, no validation error was raised in earlier versions but now it raises a 'NotSequenceError' or 'NotMappingError'.

• Rework cli string that docopt uses. Removed redundant flags that docopt provides [--version & --help]

• Add support for timestamp validation

• Add new runtime dependency 'python-dateutil' that is used to validate timestamps

• Change how 'any' keyword is implemented to now accept anything and not just the implemented types. (See Upgrade Instructions document for migration details)

Switched back to semantic version numbering for this lib.

• After the release of 15.01 the version schema was changed back from the <year>.<month> style version schema back to semantic version names. One big problem with this change is that pypi can't handle the change back to semantic names very well and because of this I had to remove the old releases from pypi and replace it with a single version 1.0.1.

• No matter what version you were using you should consider upgrading to 1.0.1. The difference between the two versions is very small and contains mostly bugfixes and added improvements.

• The old releases can still be obtained from github.com and if you really need the old version you can add the download url to your requirements.txt file.