Solr Cell isn't something i'm actively involved with – but if the issue is just having a way to pass input metadata to Tika, then perhaps a more general input params setup should be used – meta.* SolrParams that are looped over and added to the Metadata object prior to extraction perhaps?

Can it be tika.pdf.* ? The Solr parameter namespace needs some careful management.

Tika 1.1 solves TIKA-850 which will make it easier to add this feature to Solr Cell

Now that we have Tika1.1 in, we can start exploring a way to add passwords to ERH. The most flexible would perhaps be a way to specify password resolving based on regex, i.e. if the file name matches a regex for a known password you use that, if not you fallback to a default password.

For PDFs there was a possibility of supplying the password in the metadata passed on to tika (as the first patch here). However, since TIKA-850, we can now supply a PasswordProvider on the context, which will provide the password and is future proof for any document type which supports it.

Updated patch for trunk which utilizes the new Tika feature in TIKA-850. Contains a RegexRulesPasswordProvider backed by regex rules file and/or explicit password.

New solr cell request params:

• resource.password - explicit password for this file
• passwordsFile - name of property file with list of known passwords based on filename regex. Loaded using ResourceLoader

Note that Tika currently support passwords for PDF and DOCX files, not legacy DOC files or any other type. I tried to decrypt the existing test file password-is-solrcell.docx but it fails due to unsupported enctyption method in Apache POI.

In order to apply this patch and have tests pass, you also need to add two binary files by unzipping SOLR-1929-extra-docs.zip in project root.

Committed to trunk in r1354887, please check it out on a few of your own files. How to invoke:

curl "http://localhost:8983/solr/collection1/update/extract?commit=true&literal.id=123&resource.password=mypassword" \
     -H "Content-Type: application/pdf" --data-binary @my-encrypted-file.pdf

or

curl "http://localhost:8983/solr/collection1/update/extract?commit=true&literal.id=123&passwordsFile=mypass.properties&resource.name=my-encrypted-file.pdf" \
     -H "Content-Type: application/pdf" --data-binary @my-encrypted-file.pdf

# contents of mypass.properties could be:
.*\.pdf = mySecretPassword

It could of course be nice to make the PasswordProvider class pluggable through class-name as well, as we do for the CurrencyFieldType. But this is a first step and probably goes a long way.

Will keep this open until it has baked for a while in trunk and been committed to 4.x

Merged back to 4.0 beta in r1357427

Documentation updated: http://wiki.apache.org/solr/ExtractingRequestHandler#Encrypted_files

hoss20120711-manual-post-40alpha-change