The afs_spellcheck_build filter generates SpellChecker databases from PaF execution result.
The filter is declared with the afs_spellcheck_build type. It is in the antidot-paf package. It is a visitor filter.
The Spellcheck Build filter specifications are described in the following table:
Parameter name | Card. | Type | Default | Description |
auto | O | boolean | false | Toggle enabling auto_tuning of filter behavior, i.e. when document database is large then update reply databases only when necessary, and not every time. |
keep_percent | O | integer | 100 % | Percentage of document kept to build spellcheck database. |
do_nothing_if_no_document | O | boolean | false | If true, the filter will not build databases if no document has reached this filter during PaF execution. |
For all the documents of a given language, a database is generated in $AFS7/spellcheck/$LANG directory.
For documents without any language, a database is generated in $AFS7/spellcheck directory.
For each database, an about.afs file is generated. There is no global about.afs file.
This filter will only work if instanced after an afs_doc_index filter.
This filter cannot be instanced more than once. It will not read the "instances" parameter in the configuration.
This filter processes all documents of the PaF, regardless of any branch generated by a switch. In other words, it is useless to put this filter in a branch.
The spellcheck auto mode works as follows:
Number of documents | Spellchecker activation | Number of documents checked |
less than 10,000 | every run | checks 100 % of all documents |
between 10,000 and 100,000 | each 1,000 documents or 100 runs | checks 75 % of all documents |
between 100,000 and 1,000,000 | each 10,000 documents or 1,000 runs | checks 50 % of all documents |
more than 1,000,000 | each 100,000 documents or 10,000 runs | checks 10 % of all documents |