Pass MIME type and file extension to Tika

[lukasgraf]

Currently a temporary file without file extension is used to store the original document passed to Tika.

We probably should

• store the document in a temporary file using the same extension as the original file
• pass the MIME type the transform gets from the TransformEngine's convertTo method on to Tika

[lukasgraf]

It seems the Apache Tika command line interface doesn't support passing in the MIME type of the document (or any additional metadata for that matter). Tika's Detector Interface would consider such metadata, but the metadata argument seems to be only exposed in the Tika API, not the command line interface.

[lukasgraf]

So this leaves us with one option: Set the file extension of the temporary file, and let Tika's MIME type detection do its work.

The Tika Content Detection docs say that Tika

• First uses Magic bytes in the input stream
• Then refines this detection result using the file extension (if available)
• And then refines it again using the content type from the supplied metadata (which we can't set)

The command line interface help describes a switch

-d  or --detect        Detect document type

Which seems to be enabled by default (otherwise, converting a temporary file with no extension wouldn't have worked). Still, we should probably enable this switch to be sure content type detection is always performed.