Add a metadata extraction framework

Service

New MetadataService with following methods (to be improved):

• void readMetadata(DocumentModel doc)
• void writeMetadata(DocumentModel doc)
• Map<String, String> readMetadata(String processorName, Blob blob, List<String> metadataNames)
• void writeMetadata(String processorName, Blob blob, Map<String, String> metadata)
• ...

Contributions

Processor

Define a named processor, a class that will read / write metadata from/to the Blob.
Returns a Map<String, String> when reading, write a Map<String, String>.

• Map<String, String> readMetadata(Blob blob)
• void writeMetadata(Blob blob, Map<String, String> metadata)

A processor using ExifTool (will be the default one):

<extension target="org.nuxeo.metadata.MetadataService"
  point="processors">
  <processor id="exifTool" class="org.nuxeo.metadata.ExifToolProcessor" />
</extension>

Think about adding custom events to send to respond future needs of custom listener (like image magick task or else in relation to the metadata extraction)

MetadataMapping

Define a mapping between doc properties and metadata properties.

<extension target="org.nuxeo.metadata.MetadataService"
  point="metadataMappings">
  <metadataMapping id="xmp" processor="exifTool" blobXPath="file:content">
     <metadata name="tiff:ImageWidth" xpath="xmp:ImageWidth" />
     <metadata name="tiff:ImageLength" xpath="xmp:ImageLength"  />
     <metadata name="xmp:CreatorTool" xpath="xmp:CreatorTool"  />
  </metadataMapping>
</extension>

If blobXPath is empty, use BlobHolder.getBlob() to get the Blob.

Should we add a "policy" on each property to specify if read only, write or readwrite?
For now, I would ignore it and see later if we want to add it.

We should think about "merge", "override" and "append".

Rules

<extension target="org.nuxeo.metadata.MetadataService"
  point="rules">
  <rule id="default" order="0" enabled="true" async="true|false">
    <metadataMappings>
      <metadataMapping-id>xmp</metadataMapping-id>
      ...
    </metadataMappings>
    <filters>
      <filter-id>hasXMPFacet</filter-id>
    </filters>
  </rule>
</extension>

Rules apply based on the filters. Each filter is executed with the following variable in the action context:

• the document, if any
• the Blob

Default listener

Sync listener listening on creation/ modification of documents.
Trigger the MetadataService to:

• on doc creation, read the metadata and update the document (iterate on all rules) if the Blob is not empty.
• on doc modification, for each metadata mapping on each rule:
  • if Blob dirty and document metadata not dirty, read metadata from Blob to doc
  • if Blob dirty and document metadata dirty, write metadata from doc to Blob
  • if Blob not dirty and document metadata dirty, write metadata from doc to Blob

When iterating on rules, according to their "async" attribute, either call the processor to read/write, or put in the event context the dirty properties to process and then launch an event for an async listener that will process them.

Operations

At least

• Trigger a contributed metadataMapping based on its name on a document
• Write metadata to a Blob (xpath parameter, or BlobHolder if empty) from a document (input) given a custom metadata mapping defined in a Properties parameter (xpath=metadataName), using a named processor (exifTool for instance).
• Write metadata to a Blob (input) from a custom metadata mapping defined in a Properties parameter (metadataName=value), using a named processor (exifTool for instance).
• Read metadata from a Blob (xpath parameter, or BlobHolder if empty) to a document (input) given a custommetadata mapping defined in a Properties parameter (xpath=metadataName), using a named processor (exifTool for instance)
• Read metadata from a Blob (input) given a custom list of metadata defined (or optional, to get all metadata in result of ExifTool) in a StringList parameter (metadataName1, metadataName2, ...), using a named processor (exifTool for instance), and put the result (a Map) in the Context.

=> Probably need to review / add new methods on the Service to handle that.