External metadata
Introduction
External metadata is metadata which can be applied to documents in any collection type, without actually modifying the pages in any way. For example, it is possible to make all pages in a particular website match the query genre:comedy by adding a single line to the external metadata file.
www.example.org/ genre:comedy
Targets of the metadata are identified by their URL as supplied by the web server when the page was originally crawled. Note: The external metadata mechanism was designed for use only with collections containing URL data (i.e. web collections).
Metadata information can be supplied for any of the allowable metadata classes, but care should be taken if reusing reserved or special classes which have special behaviour (i.e. d
is used for dates).
Activate external metadata
To activate the external metadata for a collection, create the file external_metadata.cfg
in the collection's conf/
subdirectory. The file can be created through the Funnelback administration interface using the file-manager.
When indexing commences external_metadata.cfg
is checked for validity and data structures are set up to enable efficient lookup. If an error is detected, an appropriate error message will be printed, scanning of the file will cease and the documents will be indexed without external metadata. (See Step-Index.log
using the "Browse Log Files" option in the administration interface.)
External metadata file format
The external metadata file must be a text file delimited into lines by linefeed (\n
, hex 0x0A
) characters. Each line consists of a URL-prefix followed by a list of metadata elements which apply to all URLs which start with that prefix (unless overridden by a more specific URL-prefix).
URL prefixes must include a full hostname. It is permissible to commence the prefix with "http://". If no protocol is specified then "http://" is assumed.
Each metadata element consists of a metadata field specifier, followed by a colon, followed by a word or a string of text in double quotes. Metadata elements are separated by whitespace. Punctuation should only be present within quoted strings. e.g. genre:"Historical Drama" director:Costner
. t:"example title"
metadata will be used as the document title.
Here is an example of an external metadata file:
www.example.org publisher:"Movies Inc."
www.example.org/comedy/ genre:Comedy
www.example.org/historical-drama/ genre:"Historical Drama"
www.example.org/comedy/romance/ genre:Romance year:2012
These records have the effect that:
- Any page within the
www.example.org
site, e.g.www.example.org/movies/
orwww.example.org/about.htm
will be indexed with the metadata "publisher" = "Movies Inc." - Any page within
www.example.org/comedy/
, e.g.www.example.org/comedy/movies.htm
orwww.example.org/comedies/x/y/z.pdf
will be indexed with "genre" = "Comedy" and "publisher" = "Movies Inc." - Any page within
www.example.org/comedy/romance
e.g.www.example.org/comedy/romance/movies.htm
will be indexed with the metadata "genre" = "Romance", "year" = "2012" and "publisher" = "Movies Inc.". The "Comedy" genre will not be inherited from the second line because it was overridden in the fourth line.
Note: If multiple lines in the metadata file start with an identical prefix, only the first will be effective.
Default page handling
URL default pages (e.g. index) will be stripped from both the prefix given in the external metadata file and the URL being checked from the collection. When a URL prefix is stripped exact matching will be applied, hence example.org/index would match any default page on example.org, but would not match other pages or sub directories on example.org. Default pages are defined as any page called index, welcome, home, default or main followed by a dot and a three or four character extension (e.g. htm or html).
Metadata mapping and types
Metadata classes from external metadata don't need to have explicit metadata mappings defined and will be automatically created, however these metadata classes will be assumed to be text type, for display only. However, it is recommended that the metadata classes be defined thus explicitly setting the types for each class.
e.g. an external_metadata.cfg
might include something like
http://www.example.org/ Genre:"Comedy"
A metadata class should be set up as follows (if the class is not already defined):
- metadata class name: Genre
- metadata class type: (as appropriate)
- Source: add a custom source with a fake name, such as
fb.external_metadata
or//fb/external_metadata
The indexer actually ignores source when the metadata comes from external metadata. Only the metadata class name and type are taken into account.
External metadata date format
Dates must be specified as 8 digit integers in the format YYYYMMDD
. The document internal date is mapped to the d
metadata, any other date field will be treated as a string (unless configured otherwise in the metadata mappings):
http://www.example.org/movies/2004/ d:20040101 maxDate:20041231
This would result in the documents under http://www.example.org/movies/2004/ to have an internal date of 20040101, and an additional metadata field "expiryDate" = "20041231"
Profile-based external metadata
It is possible to create external_metadata.cfg
files per profile, rather than in the collection conf
directory, however these files won't be processed by the indexer. To have them processed they need to be concatenated and written into conf/external_metadata.cfg
prior to indexing.
This operation can be automated using the validation tool described below.
Validation and concatenation tool
mediator.pl provides a task to validate an external metadata file, as well as concatenating per-profile files into a single collection one. This task can process per-profile external metadata files, and concatenate only valid lines into the collection's conf/external_metadata.cfg
file, discarding malformed lines.
The task name is ValidateExternalMetadata
, it takes the following parameter:
collection
ID of the collection containing the external metadata file to validate or concatenate
mode
Either check
to only check an existing file for errors, or concatenate
to check and concatenate per-profile files
errorThreshold
Percentage of errors (expressed as a floating number between 0 and 1) to tolerate before exiting with an error code. For example, using errorThreshold=0.7
will tolerate up to 70% of errors in external metadata files before aborting. The default value is 0.The common use case is to configure this task in the update workflow using a pre_index_command:
pre_index_command=$SEARCH_HOME/linbin/ActivePerl/bin/perl $SEARCH_HOME/bin/mediator.pl ValidateExternalMetadata collection=$COLLECTION_NAME mode=concatenate
If there are more errors than the configured threshold, the task will fail with a non-zero exit code, causing the update to fail before indexing.
A log of the validation is produced under the collection log folder:
$SEARCH_HOME/data/<collection>/log/external_metadata.cfg-validation.log