Dev/issue 8894 premis parsing #20
Conversation
There was a problem hiding this comment.
I like the PREMIS work, but this might not be the direction we want to take PREMIS parsing.
Since this is the METS reader-writer, and METS can contain a variety of other metadata standards, it makes sense to leave the parsing of xmlData contents to something that users can easily add - some sort of plugin system. We could ship a couple of plugins for specific metadata standards ourselves (the ones we use, eg PREMIS & DublinCore), and this is a good candidate for a premis plugin, and so is https://github.com/uchicago-library/uchicagoldr-premiswork
I think this should wait to be merged until a longer term solution for how to handle parsing a variety of metadata format has been discussed, since that's out of scope for this PR.
|
|
||
| def __init__(self, title=None, creator=None, subject=None, description=None, publisher=None, contributor=None, date=None, format=None, identifier=None, source=None, relation=None, language=None, coverage=None, rights=None): | ||
| for element in self.DC_ELEMENTS: | ||
| setattr(self, element, locals()[element]) |
There was a problem hiding this comment.
This is a great use case for kwargs instead of accessing locals() directly. We could replace the list of parameters with **kwargs, which would allow us to accept an arbitrary number of inputs. kwargs is a dictionary. Then the loop could be setattr(self, element, kwargs[element])
| :raises exceptions.ParseError: If the root is not xmlData or doesn't contain a dublincore element. | ||
| """ | ||
| if root.tag != utils.lxmlns('mets') + 'xmlData': | ||
| raise exceptions.ParseError('DublinCoreXmlData can only parse xmlData elements with mets namespace.') |
There was a problem hiding this comment.
If DC is a plugin and a therefore a separate module, This should be a different, DC-specific exception, possibly that subclasses ParseError.
metsrw/metadata.py
Outdated
|
|
||
| def serialize(self): | ||
| nsmap = {'mets': utils.NAMESPACES['mets'], 'xsi': utils.NAMESPACES['xsi'], 'xlink': utils.NAMESPACES['xlink']} | ||
| root = etree.Element(utils.lxmlns('mets') + 'xmlData', nsmap=nsmap) |
There was a problem hiding this comment.
MDWrap already creates the xmlData element when serializing, and it makes sense to have DublinCore only handle elements in its own namespace. This could be merged with _serialize_dublincore below.
| parser = etree.XMLParser(remove_blank_text=True) | ||
| if isinstance(document, six.string_types): | ||
| self.document = etree.fromstring(document, parser=parser) | ||
| elif isinstance(document, etree._Element): | ||
| self.document = document | ||
| self.mdtype = mdtype | ||
| self.data = data |
There was a problem hiding this comment.
Rather than creating a new attribute, this could use document to store the child document - whether that's a string, ElementTree, or plugin class of the appropriate type.
| @@ -291,13 +345,16 @@ class MDWrap(object): | |||
| :param str mdtype: The MDTYPE of XML document being enclosed. Examples | |||
| include "PREMIS:OBJECT" and "PREMIS:EVENT". | |||
| """ | |||
| def __init__(self, document, mdtype): | |||
| MDTYPE_CLASSES = {'DC': DublinCoreXmlData} | |||
There was a problem hiding this comment.
I like this! It will have to be improved to talk to a plugin interface though.
| raise exceptions.ConstructError('event_type argument is required.') | ||
|
|
||
| if event_datetime is None: | ||
| raise exceptions.ConstructError('event_datetime argument is required.') |
There was a problem hiding this comment.
This feels unpythonic - these parameters are required by the function, and if they're None, it will raise an error later.
|
|
||
| event_type_el = etree.Element(utils.lxmlns('premis') + 'eventType') | ||
| event_type_el.text = self.event_type | ||
| root.append(event_type_el) |
There was a problem hiding this comment.
Tip: These three lines can be etree.SubElement(root, utils.lxmlns('premis') + 'eventType').text = self.event_type
|
This should be closed in favour of #34, which adds these changes to current master and makes them compatible with the recently-introduced plugin/ dependency injection pattern. Once the comments/ CR from this PR are transferred to PR 34, this can be closed. |
|
Closed in favour of #34 |
Added PREMIS object/event parsers.