Message construction and parsing
================================

This package contains helper methods to construct an RFC 2822 style message
from a list of schema fields, and to parse a message and initialise an object
based on its headers and body payload.

Before we begin, let's load the default field marshalers and configure
annotations, which we will use later in this test.

    >>> configuration = """\
    ... <configure
    ...      xmlns="http://namespaces.zope.org/zope"
    ...      i18n_domain="plone.rfc822.tests">
    ...      
    ...     <include package="zope.component" file="meta.zcml" />
    ...     <include package="zope.annotation" />
    ...     
    ...     <include package="plone.rfc822" />
    ...     
    ... </configure>
    ... """

    >>> from StringIO import StringIO
    >>> from zope.configuration import xmlconfig
    >>> xmlconfig.xmlconfig(StringIO(configuration))

The primary field
-----------------

The message body is assumed to originate from a "primary" field, which is
indicated via a marker interface.

To illustrate the pattern, consider the following schema interface:

    >>> from zope.interface import Interface, alsoProvides
    >>> from plone.rfc822.interfaces import IPrimaryField
    >>> from zope import schema

    >>> class ITestContent(Interface):
    ...
    ...     title = schema.TextLine(title=u"Title")
    ...     description = schema.Text(title=u"Description")
    ...     body = schema.Text(title=u"Body text")
    ...     emptyfield = schema.TextLine(title=u"Empty field", missing_value=u'missing')

The primary field instance is marked like this:

    >>> alsoProvides(ITestContent['body'], IPrimaryField)

Constructing a message
----------------------

Let's now say we have an instance providing this interface, which we want to
marshal to a message.

    >>> from zope.interface import implements
    >>> class TestContent(object):
    ...     implements(ITestContent)
    ...     title = u""
    ...     description = u""
    ...     body = u""
    ...     emptyfield = None

    >>> content = TestContent()
    >>> content.title = u"Test title"
    >>> content.description = u"""Test description
    ... with a newline"""
    >>> content.body = u"<p>Test body</p>"

We could create a message form this instance and schema like this:

    >>> from plone.rfc822 import constructMessageFromSchema
    >>> msg = constructMessageFromSchema(content, ITestContent)

The output looks like this:

    >>> from plone.rfc822 import renderMessage
    >>> print renderMessage(msg)
    title: Test title
    description: =?utf-8?q?Test_description=0D=0Awith_a_newline?=
    emptyfield: 
    Content-Type: text/plain; charset="utf-8"
    <BLANKLINE>
    <p>Test body</p>

Notice how the non-ASCII header values are UTF-8 encoded. The encoding
algorithm is clever enough to only encode the value if it is necessary,
leaving more readable field values otherwise.

The body here is of the default message type:

    >>> msg.get_default_type()
    'text/plain'

This is because none of the default field types manage a content type.

The body is also utf-8 encoded, because the primary field specified this
encoding.

If we want to use a different content type, we could set it explicitly:

    >>> msg.set_type('text/html')
    >>> print renderMessage(msg)
    title: Test title
    description: =?utf-8?q?Test_description=0D=0Awith_a_newline?=
    emptyfield: 
    MIME-Version: 1.0
    Content-Type: text/html; charset="utf-8"
    <BLANKLINE>
    <p>Test body</p>

Alternatively, if we know that any ``IText`` field on an object providing
our ``ITestContent`` interface always stores HTML, could register a custom
``IFieldMarshaler`` adapter which would indicate this to the message
constructor. Let's take a look at that now.

Custom marshalers
-----------------

The default marshaler can be obtained by multi-adapting the content object
and the field instance to ``IFieldMarshaler``:

    >>> from zope.component import getMultiAdapter
    >>> from plone.rfc822.interfaces import IFieldMarshaler
    >>> getMultiAdapter((content, ITestContent['body'],), IFieldMarshaler)
    <plone.rfc822.defaultfields.UnicodeValueFieldMarshaler object at ...>

Let's now create our own marshaler by extending this class and overriding
the ``getContentType()``:

    >>> from plone.rfc822.defaultfields import UnicodeValueFieldMarshaler
    >>> from zope.schema.interfaces import IText
    >>> from zope.component import adapts

    >>> class TestBodyMarshaler(UnicodeValueFieldMarshaler):
    ...     adapts(ITestContent, IText)
    ...     
    ...     def getContentType(self):
    ...         return 'text/html'

Ordinarily, we'd register this with ZCML. For the purpose of the test, we'll
register it using the ``zope.component`` API.

    >>> from zope.component import provideAdapter
    >>> provideAdapter(TestBodyMarshaler)

Hint: If the schema contained multiple text fields, this adapter would apply
to all of them. To avoid that, we could either mark the field with a custom
marker interface (similary to the way we marked a field with ``IPrimaryField``
above), or have the marshaler check the field name.

Let's now try again:

    >>> msg = constructMessageFromSchema(content, ITestContent)
    >>> print renderMessage(msg)
    title: Test title
    description: =?utf-8?q?Test_description=0D=0Awith_a_newline?=
    emptyfield: 
    MIME-Version: 1.0
    Content-Type: text/html; charset="utf-8"
    <BLANKLINE>
    <p>Test body</p>

Notice how the Content-Type has changed.

Consuming a message
-------------------

A message can be used to initialise an object. The object has to be
constructed first:

    >>> newContent = TestContent()

We then need to obtain a ``Message`` object. The ``email`` module contains
helper functions for this purpose.

    >>> messageBody = """\
    ... title: Test title
    ... description: =?utf-8?q?Test_description=0D=0Awith_a_newline?=
    ... Content-Type: text/html
    ...
    ... <p>Test body</p>"""

    >>> from email import message_from_string
    >>> msg = message_from_string(messageBody)

The message can now be used to initialise the object according to the given
schema. This should be the same schema as the one used to construct the
message.

    >>> from plone.rfc822 import initializeObjectFromSchema
    >>> initializeObjectFromSchema(newContent, ITestContent, msg)

    >>> newContent.title
    u'Test title'
    >>> newContent.description
    u'Test description\nwith a newline'
    >>> newContent.body
    u'<p>Test body</p>'

We can also consume messages with a transfer encoding and a charset:

    >>> messageBody = """\
    ... title: =?utf-8?q?Test_title?=
    ... description: =?utf-8?q?Test_description=0D=0Awith_a_newline?=
    ... emptyfield: 
    ... Content-Transfer-Encoding: base64
    ... Content-Type: text/html; charset="utf-8"
    ... <BLANKLINE>
    ... PHA+VGVzdCBib2R5PC9wPg==
    ... <BLANKLINE>"""

    >>> msg = message_from_string(messageBody)
    >>> msg.get_content_type()
    'text/html'
    >>> msg.get_content_charset()
    'utf-8'
    
    >>> initializeObjectFromSchema(newContent, ITestContent, msg)

    >>> newContent.title
    u'Test title'
    >>> newContent.description
    u'Test description\nwith a newline'
    >>> newContent.body
    u'<p>Test body</p>'

Note: Empty fields will result in the field's ``missing_value`` being used:

    >>> newContent.emptyfield
    u'missing'

Handling multiple primary fields and duplicate field names
----------------------------------------------------------

It is possible that our type could have multiple primary fields or even
duplicate field names.

For example, consider the following schema interface, intended to be used
in an annotation adapter:

    >>> class IPersonalDetails(Interface):
    ...     description = schema.Text(title=u"Personal description")
    ...     currentAge = schema.Int(title=u"Age", min=0)
    ...     personalProfile = schema.Text(title=u"Profile")
    
    >>> alsoProvides(IPersonalDetails['personalProfile'], IPrimaryField)
    
The annotation storage would look like this:

    >>> from persistent import Persistent
    >>> class PersonalDetailsAnnotation(Persistent):
    ...     implements(IPersonalDetails)
    ...     adapts(ITestContent)
    ...     
    ...     def __init__(self):
    ...         self.description = None
    ...         self.currentAge = None
    ...         self.personalProfile = None

    >>> from zope.annotation.factory import factory
    >>> provideAdapter(factory(PersonalDetailsAnnotation))

We should now be able to adapt a content instance to IPersonalDetails,
provided it is annotatable.

    >>> from zope.annotation.interfaces import IAttributeAnnotatable
    >>> alsoProvides(content, IAttributeAnnotatable)

    >>> personalDetails = IPersonalDetails(content)
    >>> personalDetails.description = u"<p>My description</p>"
    >>> personalDetails.currentAge = 21
    >>> personalDetails.personalProfile = u"<p>My profile</p>"

The default marshalers will attempt to adapt the context to the schema of
a given field before getting or setting a value. If we pass multiple schemata
(or a combined sequence of fields) to the message constructor, it will
handle both duplicate field names (as duplicate headers) and multiple primary
fields (as multipart message attachments).

Here are the fields it will see:

    >>> from zope.schema import getFieldsInOrder
    >>> allFields = getFieldsInOrder(ITestContent) + \
    ...             getFieldsInOrder(IPersonalDetails)

    >>> [f[0] for f in allFields]
    ['title', 'description', 'body', 'emptyfield', 'description', 'currentAge', 'personalProfile']

    >>> [f[0] for f in allFields if IPrimaryField.providedBy(f[1])]
    ['body', 'personalProfile']

Let's now construct a message. Since we now have two fields called 
``description``, we will get two headers by that name. Since we have two
primary fields, we will get a multipart message with two attachments.

    >>> from plone.rfc822 import constructMessageFromSchemata
    >>> msg = constructMessageFromSchemata(content, (ITestContent, IPersonalDetails,))
    >>> msgString = renderMessage(msg)
    >>> print msgString
    title: Test title
    description: =?utf-8?q?Test_description=0D=0Awith_a_newline?=
    emptyfield: 
    description: <p>My description</p>
    currentAge: 21
    MIME-Version: 1.0
    Content-Type: multipart/mixed; boundary="===============...=="
    <BLANKLINE>
    --===============...==
    MIME-Version: 1.0
    Content-Type: text/html; charset="utf-8"
    <BLANKLINE>
    <p>Test body</p>
    --===============...==
    MIME-Version: 1.0
    Content-Type: text/html; charset="utf-8"
    <BLANKLINE>
    <p>My profile</p>
    --===============...==--...

(Note that we've used ellipses here for the doctest to work with the generated
boundary string).

Notice how both messages have a MIME type of 'text/html' and no charset.
That is because of the custom adapter for ``(ITestContent, IText)`` which we
registered earlier.

We can obviously read this message as well. Note that in this case, the order
of fields passed to ``initializeObject()`` is important, both to determine
which field gets which ``description`` header, and to match the two
attachments to the two primary fields:

    >>> newContent = TestContent()
    >>> alsoProvides(newContent, IAttributeAnnotatable)

    >>> from plone.rfc822 import initializeObjectFromSchemata
    >>> msg = message_from_string(msgString)
    >>> initializeObjectFromSchemata(newContent, [ITestContent, IPersonalDetails], msg)

    >>> newContent.title
    u'Test title'
    
    >>> newContent.description
    u'Test description\nwith a newline'

    >>> newContent.body
    u'<p>Test body</p>'

    >>> newPersonalDetails = IPersonalDetails(newContent)
    >>> newPersonalDetails.description
    u'<p>My description</p>'

    >>> newPersonalDetails.currentAge
    21
    
    >>> newPersonalDetails.personalProfile
    u'<p>My profile</p>'
    
Alternative ways to deal with multiple schemata
-----------------------------------------------

In the example above, we created a single enveloping message with headers
corresponding to the fields in both our schemata, and only the primary fields
separated out into different attached payloads.

An alternative approach would be to separate each schema out into its
own multipart message. To do that, we would simply use the
``constructMessage()`` function multiple times.

    >>> mainMessage = constructMessageFromSchema(content, ITestContent)
    >>> personalDetailsMessage = constructMessageFromSchema(content, IPersonalDetails)

    >>> from email.MIMEMultipart import MIMEMultipart
    >>> envelope = MIMEMultipart()
    >>> envelope.attach(mainMessage)
    >>> envelope.attach(personalDetailsMessage)

    >>> envelopeString = renderMessage(envelope)
    >>> print envelopeString
    Content-Type: multipart/mixed; boundary="===============...=="
    MIME-Version: 1.0
    <BLANKLINE>
    --===============...==
    title: Test title
    description: =?utf-8?q?Test_description=0D=0Awith_a_newline?=
    emptyfield: 
    MIME-Version: 1.0
    Content-Type: text/html; charset="utf-8"
    <BLANKLINE>
    <p>Test body</p>
    --===============...==
    description: <p>My description</p>
    currentAge: 21
    MIME-Version: 1.0
    Content-Type: text/html; charset="utf-8"
    <BLANKLINE>
    <p>My profile</p>
    --===============...==--...

Which approach works best will depend largely on the intended recipient of
the message.

Encoding the payload and handling filenames
-------------------------------------------

Finally, let's consider a more complex example, inspired by the field
marshaler in ``plone.namedfile``.

Let's say we have a value type intended to represent a binary file with a
filename and content type:

    >>> from zope.interface import Interface, implements
    >>> from zope import schema
    
    >>> class IFileValue(Interface):
    ...     data = schema.Bytes(title=u"Raw data")
    ...     contentType = schema.ASCIILine(title=u"MIME type")
    ...     filename = schema.ASCIILine(title=u"Filename")

    >>> class FileValue(object):
    ...     implements(IFileValue)
    ...     def __init__(self, data, contentType, filename):
    ...         self.data = data
    ...         self.contentType = contentType
    ...         self.filename = filename

Suppose we had a custom field type to represent this:

    >>> from zope.schema.interfaces import IObject
    >>> class IFileField(IObject):
    ...     pass

    >>> class FileField(schema.Object):
    ...     implements(IFileField)
    ...     schema = IFileValue
    ...     def __init__(self, **kw):
    ...         if 'schema' in kw:
    ...             self.schema = kw.pop('schema')
    ...         super(FileField, self).__init__(schema=self.schema, **kw)

We can register a field marshaler for this field which will do the following:

* Insist that the field is only used as a primary field, since it makes
  little sense to encode a binary file in a header.
* Save the filename in a Content-Disposition header.
* Be capable of reading the filename again from this header.
* Encode the payload using base64

    >>> from plone.rfc822.interfaces import IFieldMarshaler
    >>> from email.Encoders import encode_base64

    >>> from zope.component import adapts
    >>> from plone.rfc822.defaultfields import BaseFieldMarshaler
    
    >>> class FileFieldMarshaler(BaseFieldMarshaler):
    ...     adapts(Interface, IFileField)
    ...
    ...     ascii = False
    ...
    ...     def encode(self, value, charset='utf-8', primary=False):
    ...         if not primary:
    ...             raise ValueError("File field cannot be marshaled as a non-primary field")
    ...         if value is None:
    ...             return None
    ...         return value.data
    ...     
    ...     def decode(self, value, message=None, charset='utf-8', contentType=None, primary=False):
    ...         filename = None
    ...         # get the filename from the Content-Disposition header if possible
    ...         if primary and message is not None:
    ...             filename = message.get_filename(None)
    ...         return FileValue(value, contentType, filename)
    ...     
    ...     def getContentType(self):
    ...         value = self._query()
    ...         if value is None:
    ...             return None
    ...         return value.contentType
    ...
    ...     def getCharset(self, default='utf-8'):
    ...         return None # this is not text data!
    ...
    ...     def postProcessMessage(self, message):
    ...         value = self._query()
    ...         if value is not None:
    ...             filename = value.filename
    ...             if filename:
    ...                 # Add a new header storing the filename if we have one
    ...                 message.add_header('Content-Disposition', 'attachment', filename=filename)
    ...         # Apply base64 encoding
    ...         encode_base64(message)
    
    >>> from zope.component import provideAdapter
    >>> provideAdapter(FileFieldMarshaler)

To illustrate marshaling, let's create a content object that contains two file
fields.
    
    >>> class IFileContent(Interface):
    ...     file1 = FileField()
    ...     file2 = FileField()

    >>> class FileContent(object):
    ...     implements(IFileContent)
    ...     file1 = None
    ...     file2 = None

    >>> fileContent = FileContent()
    >>> fileContent.file1 = FileValue('dummy file', 'text/plain', 'dummy1.txt')
    >>> fileContent.file2 = FileValue('<html><body>test</body></html>', 'text/html', 'dummy2.html')

At this point, neither of these fields is marked as a primary field. Let's see
what happens when we attempt to construct a message from this schema.

    >>> from plone.rfc822 import constructMessageFromSchema
    >>> message = constructMessageFromSchema(fileContent, IFileContent)
    >>> print renderMessage(message)
    <BLANKLINE>
    <BLANKLINE>

As expected, we got no message headers and no message body. Let's now mark one
field as primary:

    >>> from plone.rfc822.interfaces import IPrimaryField
    >>> from zope.interface import alsoProvides
    >>> alsoProvides(IFileContent['file1'], IPrimaryField)
    
    >>> message = constructMessageFromSchema(fileContent, IFileContent)
    >>> messageBody = renderMessage(message)
    >>> print messageBody
    MIME-Version: 1.0
    Content-Type: text/plain
    Content-Disposition: attachment; filename="dummy1.txt"
    Content-Transfer-Encoding: base64
    <BLANKLINE>
    ZHVtbXkgZmlsZQ==

Here, we have a base64 encoded payload, a Content-Disposition header, and a
Content-Type header according to the primary field.

We can also reconstruct the object from this message.

    >>> from plone.rfc822 import initializeObjectFromSchema
    >>> from email import message_from_string
    
    >>> inputMessage = message_from_string(messageBody)
    >>> newFileContent = FileContent()
    >>> initializeObjectFromSchema(newFileContent, IFileContent, inputMessage)
    
    >>> newFileContent.file1.data
    'dummy file'
    >>> newFileContent.file1.contentType
    'text/plain'
    >>> newFileContent.file1.filename
    'dummy1.txt'

    >>> newFileContent.file2 is None
    True

Let's now show what would happen if we encoded both files in the message.
In this case, we should get a multipart document with two payloads.

    >>> alsoProvides(IFileContent['file2'], IPrimaryField)
    >>> message = constructMessageFromSchema(fileContent, IFileContent)
    >>> messageBody = renderMessage(message)
    >>> print messageBody # doctest: +ELLIPSIS
    MIME-Version: 1.0
    Content-Type: multipart/mixed; boundary="===============...=="
    <BLANKLINE>
    --===============...==
    MIME-Version: 1.0
    Content-Type: text/plain
    Content-Disposition: attachment; filename="dummy1.txt"
    Content-Transfer-Encoding: base64
    <BLANKLINE>
    ZHVtbXkgZmlsZQ==
    --===============...==
    MIME-Version: 1.0
    Content-Type: text/html
    Content-Disposition: attachment; filename="dummy2.html"
    Content-Transfer-Encoding: base64
    <BLANKLINE>
    PGh0bWw+PGJvZHk+dGVzdDwvYm9keT48L2h0bWw+
    --===============...==--...

And again, we can reconstruct the object, this time with both fields:

    >>> inputMessage = message_from_string(messageBody)
    >>> newFileContent = FileContent()
    >>> initializeObjectFromSchema(newFileContent, IFileContent, inputMessage)
    
    >>> newFileContent.file1.data
    'dummy file'
    >>> newFileContent.file1.contentType
    'text/plain'
    >>> newFileContent.file1.filename
    'dummy1.txt'

    >>> newFileContent.file2.data
    '<html><body>test</body></html>'
    >>> newFileContent.file2.contentType
    'text/html'
    >>> newFileContent.file2.filename
    'dummy2.html'