A simple schema-based serialization and deserialization library
Project description
Colander
An extensible package which can be used to:
deserialize and validate a data structure composed of strings, mappings, and lists.
serialize an arbitrary data structure to a data structure composed of strings, mappings, and lists.
Please see http://docs.pylonsproject.org/projects/colander/dev/ for further documentation.
Changes
0.9.4 (2011-10-14)
flatten now only includes leaf nodes in the flattened dict.
flatten does not include a path element for the name of the type node for sequences.
unflatten is implemented.
Added __setitem__ to SchemaNode, allowing replacement of nodes by name.
Added get_value and set_value methods to Schema which allow access and mutation of appstructs using dotted name paths.
Add Swedish, French, Chinese translations.
0.9.3 (2011-06-23)
Add Time type.
Add Dutch translation.
Fix documentation: 0.9.2 requires deserialize of types to explicitly deal with the potential to receive colander.null.
Use default_tzinfo when deserializing naive datetimes. See https://github.com/Pylons/colander/issues#issue/5.
Allow default_tzinfo to be None when creating a colander.DateTime. See https://github.com/Pylons/colander/issues#issue/6.
Add the ability to insert a colander.interfaces.Preparer between deserialization and validation. See the Preparing section in the documentation.
0.9.2 (2011-03-28)
Added Polish translation, thanks to Jedrzej Nowak.
Moved to Pylons Project GitHub (https://github.com/Pylons/colander).
Add tox.ini for testing purposes.
New API: colander.required. Used as the marker value when a missing argument is left unspecified.
Bug fix: if a title argument which is the empty string or None is passed explicitly to a SchemaNode, it is no longer replaced by a title computed from the name.
Add SchemaNode.__contains__ to support “name in schema”.
SchemaNode deserialization now unconditionally calls the schema type’s deserialize method to obtain an appstruct before attempting to validate. Third party schema types should now return colander.null if passed a colander.null value or another logically “empty” value as a cstruct during deserialize.
0.9.1 (2010-12-02)
When colander.null was unpickled, the reference created during unpickling was not a reference to the singleton but rather a new instance of the colander._null class. This was unintentional, because lots of code checks for if x is colander.null, which will fail across pickling and unpickling. Now the reference created when colander.null is pickled is unpickled as the singleton itself.
0.9 (2010-11-28)
SchemaNode constructor now accepts arbitrary keyword arguments. It sets any unknown values within the **kw sequence as attributes of the node object.
Added Spanish locale: thanks to Douglas Cerna for the translations!
If you use a schema with deferred validator, missing or default attributes, but you use it to perform serialization and deserialization without calling its bind method:
If validator is deferred, no validation will be performed.
If missing is deferred, the field will be considered required.
If default is deferred, the serialization default will be assumed to be colander.null.
Undocumented internal API for all type objects: flatten. External type objects should now inherit from colander.SchemaType to get a default implementation.
0.8 (2010/09/08)
Docstring fixes to colander.SchemaNode (missing is not the null value when required, it’s a special marker value).
The concept of “schema binding” was added, which allows for a more declarative-looking spelling of schemas and schema nodes which have dependencies on values available after the schema has already been fully constructed. See the new narrative chapter in the documentation entitled “Schema Binding”.
The interface of colander.SchemaNode has grown a __delitem__ method. The __iter__, and __getitem__ methods have now also been properly documented.
0.7.3 (2010/09/02)
The title of a schema node now defaults to a titleization of the name. Underscores in the name are replaced with empty strings and the first letter of every resulting word is capitalized. Previously the name was not split on underscores, and the entirety of the name was capitalized.
A method of the colander.Invalid exception named messages was added. It returns an iterable of error messages using the msg attribute of its related exception node. If the msg attribute is iterable, it is returned. If it is not iterable, a single-element list containing the msg value is returned.
0.7.2 (2010/08/30)
Add an colander.SchemaNode.__iter__ method, which iterates over the children nodes of a schema node.
The constructor of a colander.SchemaNode now accepts a widget keyword argument, for use by Deform (it is not used internally).
0.7.1 (2010/06/12)
Make it possible to use colander.null as a missing argument to colander.SchemaNode for roundtripping purposes.
Make it possible to pickle colander.null.
0.7.0
A release centered around normalizing the treatment of default and missing values.
Bug Fixes
Allow colander.Regex validator to accept a pattern object instead of just a string.
Get rid of circular reference in Invalid exceptions: Invalid exceptions now no longer have a parent attribute. Instead, they have a positional attribute, which signifies that the parent node type of the schema node to which they relate inherits from Positional. This attribute isn’t an API; it’s used only internally for reporting.
Raise a TypeError when bogus keyword arguments are passed to colander.SchemaNode.
Backwards Incompatiblities / New Features
missing constructor arg to SchemaNode: signifies deserialization default, disambiguated from default which acted as both serialization and deserialization default previously.
Changes necessitated / made possible by SchemaNode missing addition:
The allow_empty argument of the colander.String type was removed (use missing='' as a wrapper SchemaNode argument instead).
New concept: colander.null input to serialization and deserialization. Use of colander.null normalizes serialization and deserialization default handling.
Changes necessitated / made possible by colander.null addition:
- partial argument and attribute of colander.MappingSchema has
been removed; all serializations are partial, and partial deserializations are not necessary.
- colander.null values are added to the cstruct for partial
serializations instead of omitting missing node values from the cstruct.
- colander.null may now be present in serialized and
deserialized data structures.
sdefault attribute of SchemaNode has been removed; we never need to serialize a default anymore.
The value colander.null will be passed as appstruct to each type’s serialize method when a mapping appstruct doesn’t have a corresponding key instead of None, as was the practice previously.
The value colander.null will be passed as cstruct to each type’s deserialize method when a mapping cstruct doesn’t have a corresponding key instead of None, as was the practice previously.
Types now must handle colander.null explicitly during serialization.
Updated and expanded documentation, particularly with respect to new colander.null handling.
The value argument`` to the serialize method of a SchemaNode is now named appstruct. It is no longer a required argument; it defaults to colander.null now.
The value argument to the deserialize method of a SchemaNode is now named cstruct. It is no longer a required argument; it defaults to colander.null now.
The value argument to the serialize method of each built-in type is now named appstruct, and is now required: it is no longer a keyword argument that has a default.
The value argument to the deserialize method of each built-in type is now named cstruct, and is now required: it is no longer a keyword argument that has a default.
0.6.2 (2010-05-08)
The default encoding parameter value to the colander.String type is still None, however its meaning has changed. An encoding of None now means that no special encoding and decoding of Unicode values is done by the String type. This differs from the previous behavior, where None implied that the encoding was utf-8. Pass the encoding as utf-8 specifically to get the older behavior back. This is in support of Deform.
The default err_template value attached to the colander.Date and colander.Datetime types was changed. It is now simply Invalid date instead of _('${val} cannot be parsed as an iso8601 date: ${err}'). This is in support of Deform.
Fix bug in colander.Boolean that attempted to call .lower on a bool value when a default value was found for the schema node.
0.6.1 (2010-05-04)
Add a Decimal type (number type which uses decimal.Decimal as a deserialization target).
0.6.0 (2010-05-02)
(Hopefully) fix intermittent datetime-granularity-related test failures.
Internationalized error messages. This required some changes to error message formatting, which may impact you if you were feeding colander an error message template.
New project dependency: translationstring package for internationalization.
New argument to colander.String constructor: allow_empty. This is a boolean representing whether an empty string is a valid value during deserialization, defaulting to False.
Add minimal documentation about the composition of a colander.Invalid exception to the narrative docs.
Add (existing, but previously non-API) colander.Invalid attributes to its interface within the API documentation.
0.5.2 (2010-04-09)
Add Email and Regex validators (courtesy Steve Howe).
Raise a colander.Invalid error if a colander.SequenceSchema is created with more than one member.
Add Function validator.
Fix bug in serialization of non-Unicode values in the String class.
Get rid of pserialize in favor of making serialize always partially serialize.
Get rid of pdeserialize: it existed only for symmetry. We’ll add something like it back later if we need it.
0.5.1 (2010-04-02)
The constructor arguments to a the colander.Schema class are now sent to the constructed SchemaNode rather than to the type it represents.
Allow colander.Date and colander.DateTime invalid error messages to be customized.
Add a pos argument to the colander.Invalid.add method.
Add a __setitem__ method to the colander.Invalid class.
The colander.Mapping constructor keyword argument unknown_keys has been renamed to unknown.
Allow colander.Mapping type to accept a new constructor argument: partial.
New interface methods required by types and schema nodes: pserialize and pdeserialize. These partially serialize or partially deserialize a value (the definition of “partial” is up to the type).
0.5 (2010-03-31)
0.4 was mispackaged (CHANGES.txt missing); no code changes from 0.4 however.
0.4 (2010-03-30)
Add colander.DateTime and colander.Date data types.
Depend on the iso8601 package for date support.
0.3 (2010-03-29)
Subnodes of a schema node are now kept in the children attribute rather than the nodes attribute.
Add an sdefault property to colander.SchemaNode objects.
Add a clone method to colander.SchemaNode objects.
Add a __str__ method to the colander.Invalid exception that prints an error summary.
Various error message improvements.
Add colander.Length validator class.
0.2 (2010-03-23)
Make nodetype overrideable.
Add __getitem__ to SchemaNode.
Fix OneOf message.
Capitalize node titles.
Deal with empty strings in String, Boolean, and Float types.
Introduce description; make title the same as name.
Remove copy method from schemanode.
Allow schema nodes to have titles.
The term “structure” is too overloaded to use for schema purposes: structure -> schema node.
Make Sequence more like Tuple and Mapping (it uses a substructure rather than a structure parameter to denote its type).
Add __repr__ and copy methods to structure.
Add accept_scalar flag to Sequence.
0.1 (2010-03-14)
Initial release.
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
