man SWISH::API () - Perl interface to the Swish-e C Library

NAME

SWISH::API - Perl interface to the Swish-e C Library

SYNOPSIS

    use SWISH::API;

    my $swish = SWISH::API->new( 'index.swish-e' );

    $swish->AbortLastError
        if $swish->Error;

    # A short-cut way to search

    my $results = $swish->Query( "foo OR bar" );

    # Or more typically
    my $search = $swish->New_Search_Object;

    # then in a loop
    my $results = $search->Execute( $query );

    # always check for errors (but aborting is not always necessary)

    $swish->AbortLastError
        if $swish->Error;

    # Display a list of results

    my $hits = $results->Hits;
    if ( !$hits ) {
        print "No Results\n";
        return;  /* for example *.
    }

    print "Found ", $results->Hits, " hits\n";

    # Seek to a given page - should check for errors
    $results->SeekResult( ($page-1) * $page_size );

    while ( my $result = $results->NextResult ) {
        printf("Path: %s\n  Rank: %lu\n  Size: %lu\n  Title: %s\n  Index: %s\n  Modified: %s\n  Record #: %lu\n  File   #: %lu\n\n",
            $result->Property( "swishdocpath" ),
            $result->Property( "swishrank" ),
            $result->Property( "swishdocsize" ),
            $result->Property( "swishtitle" ),
            $result->Property( "swishdbfile" ),
            $result->ResultPropertyStr( "swishlastmodified" ),
            $result->Property( "swishreccount" ),
            $result->Property( "swishfilenum" )
        );
    }

    # display properties and metanames

    for my $index_name ( $swish->IndexNames ) {
        my @metas = $swish->MetaList( $index_name );
        my @props = $swish->PropertyList( $index_name );

        for my $m ( @metas ) {
            my $name = $m->Name;
            my $id = $m->ID;
            my $type = $m->Type;
        }
        # (repeat above for @props)
    }

DESCRIPTION

This module provides a Perl interface to the Swish-e search engine. This module allows embedding the swish-e search code into your application avoiding the need to fork to run the swish-e binary and to keep an index file open when running multiple queries. This results in increased search performance.

DEPENDENCIES

You must have installed Swish-e version 2.4 before building this module. Download from:

    http://swish-e.org

OVERVIEW

This module includes a number of classes.

Searching consists of connecting to a swish-e index (or indexes), and then running queries against the open index. Connecting to the index creates a swish object blessed into the SWISH::API class.

A SWISH::API::Search object is created from the SWISH::API object. The SWISH::API::Search object can have associated parameters (e.g. result sort order).

The SWISH::API::Search object is used to query the associated index file or files. A query on a search object returns a results object of the class SWISH::API::Results. Then individual results of the SWISH::API::Result class can be fetched by calling a method of the results object.

Finally, a result's properties can be accessed by calling methods on the result object.

METHODS

SWISH::API - Swish Handle Object

To begin using Swish you must first create a Swish Handle object. This object makes the connection to one or more index files and is used to create objects used for searching the associated index files. This method returns a swish handle object blessed into the SWISH::API class. CW$index_files is a space separated list of index files to open. This always returns an object, even on errors. Caller must check for errors (see below). Returns a list of index names associated with the swish handle. These were the indexes specified as a parameter on the SWISH::API->new call. This can be used in calls below that require specifying the index file name. Returns a list of possible header names. These can be used to lookup header values. See CWSwishHeaderValue method below. A swish-e index has data associated with it stored in the index header. This method provides access to that data. Returns the header value for the header and index file specified. Most headers are a single item, but some headers (e.g. Stopwords) return a list. The list of possible header names can be obtained from the SwishHeaderNames method.

$swish->RankScheme( 0|1 );
Similar to the -R option with the swish-e command line tool. The default ranking scheme is 0. Set it to 1 to experiment with other ranking features. See the SWISH-CONFIG documentation for more on ranking schemes.

Error Handling

All errors are stored in and accessed via the SWISH::API object (the Swish Handle). That is, even an error that occurs when calling a method on a result (SWISH::API::Result) object will store the error in the parent SWISH:API object.

Check for errors after every method call. Some errors are critical errors and will require destruction of the SWISH::API object. Critical errors will typically only happen when attaching to the database and are errors such as an invalid index file name, permissions errors, or passing invalid objects to calls.

Typically, if you receive an error when attaching to an index file or files you should assume that the error is critical and let the swish object fall out of scope (and destroyed). Otherwise, if an error is detected you should check if it is a critical error. If the error is not critical you may continue using the objects that have been created (for example, an invalid meta name will generate a non-critical error, so you may continue searching using the same search object).

Error state is cleared upon a new query.

Again, all error methods need to be called on the parent swish object

$swish->Error
Returns true if an error occurred on the last operation. On errors the value returned is the internal Swish-e error number (which is less than zero).
$swish->CriticalError
Returns true if the last error was a critical error
$swish->AbortLastError
Aborts the running program and prints an error message to STDERR. Returns the string description of the current error (based on the value returned by CW$swish->Error). This is a generic error string. Returns a string with specific information about the last error, if any. For example, if a query of:
    badmeta=foo
and badmeta is an invalid metaname CW$swish->ErrorString might return Unknown metaname, but CW$swish->LastErrorMsg might return badmeta.

Generating Search and Result Objects This creates a new search object blessed into the SWISH::API::Search class. The optional CW$query parameter is a query string to store in the search object. See the section on CWSWISH::API::Search for methods available on the returned object. The advantage of this method is that a search object can be used for multiple queries:

    $search = $swish->New_Search_Objet;
    while ( $query = next_query() ) {
        $results = $search->Execute( $query );
        ...
    }
This is a short-cut which avoids the step of creating a separate search object. It returns a results object blessed into the SWISH::API::Results class described below. This method basically is the equivalent of
    $results = $swish->New_Search_Object->Execute( $query );

SWISH::API::Search - Search Objects

A search object holds the parameters used to generate a list of results. These methods are used to adjust these parameters and to create the list of results for the current set of search parameters. This will set (or replace) the query string associated with a search object. This method is typically not used as the query can be set when executing the actual query or when creating a search object. This method may change in the future. A structure is a bit-mapped flag used to limit search results to specific parts of an HTML document, such as the title or in H tags. The possible bits are:

    IN_FILE         = 1      This is the default
    IN_TITLE        = 2      In <title> tag
    IN_HEAD         = 4      In <head> tag
    IN_BODY         = 8      In <body>
    IN_COMMENTS     = 16     In html comments
    IN_HEADER       = 32     In <h*>
    IN_EMPHASIZED   = 64     In <em>, <b>, <strong>, <i>
    IN_META         = 128    In a meta tag (e.g. not swishdefault)
So if you wish to limit your searches to words in heading tags (e.g. <H1>) or in the <title> tag use:
    $search->SetStructure( IN_HEAD | IN_TITLE );
Sets the character used as the phrase delimiter in searches. The default is double-quotes ("). Sets a range from CW$low to CW$high inclusive that the give CW$property must be in to be selected as a result. Call multiple times to set more than one limit on different properties. Limits are ANDed, that is, a result must be within the range of all limits specified to be included in a list of results. For example to limit searches to documents modified in the last 48 hours:
    my $start = time - 48 * 60 * 60;
    $search->SetSearchLimit( 'swishlastmodified', $start, time() );
An error will be set if the property has already been specified or if CW$high < CW$low. Other errors may not be reported until running the query, such as the property name is invalid or if CW$low or CW$high are not numeric and the property specified is a numeric property. Once a query is run you cannot change the limit settings without calling the ResetSearchLimit method first.
$search->ResetSearchLimit;
Clears the limit parameters for the given object. This must be called if the limit parameters need to be changed. Sets the sort order of search results. The string is a space separated list of valid document properties. Each property may contain a qualifier that sets the direction of the sort. For example, to sort the results by path name in ascending order and by rank in descending order:
    $search->SetSort( 'swishdocpath asc swishrank desc' );
The asc and desc qualifiers are optional, and if omitted ascending is assumed. Currently, errors (e.g invalid property name) are not detected on this call, but rather when executing a query. This may change in the future.

SWISH::API::Results - Generating and accessing results

Searching generates a results object blessed into the SWISH::API::Results class. Executes a query based on the parameters in the search object. CW$query is an optional query string to use for the search ($query replaces the set query string in the search object). A typical use would be to create a search object once and then call this method for each query using the same search object changing only the passed in CW$query. The caller should check for errors after making this all.

Results Methods

A query creates a results object that contains information about the query (e.g. number of hits) and access to the individual results. Returns the number of results for the query. If zero and no errors were reported after calling CW$search->Execute then the query returned zero results. Returns an array of tokenized words and operators with stopwords removed. This is the array of tokens used by swish for the query. $index_name must match one of the index files specified on the creation of the swish object (via the SWISH::API->new call). The parsed words are useful for highlighting search terms in associated documents. Returns an array of stopwords removed from a query, if any, for the index specified. $index_name must match one of the index files specified on the creation of the swish object (via the SWISH::API->new call). Seeks to the position specified in the result list. Zero is the first position and CW$results->Hits-1 is the last position. Seeking past the end of results sets a non-critical error condition. Useful for seeking to a specific page of results. Fetches the next result from the list of results. Returns undef if no more results are available. CW$result is an object blessed into the SWISH::API::Result class.

SWISH::API::Result - Result Methods

The follow methods provide access to data related to an individual result. Fetches the property specified for the current result. An invalid property name will cause an exception (which can be caught by wrapping the call in an eval block). Can return undefined. Date properties are returned as a timestamp. Use something like Date::Format to format the strings (or just call scalar localtime( CW$prop ) ). Fetches and formats the property. Unlike above, invalid property names return the string (null) this will likely change to match the above (i.e. throw an exception). Undefined values are returned at the null string (""). Returns the header value specified. This is similar to CW$swish->HeaderValue(), but the index file is not specified (it is determined by the result).

Utility Methods

Swish-e has MetaNames which allow searching by fields in the index. This method returns information about the Metanames. Pass in the name of an open index file name and returns a list of SWISH::API::MetaName objects. Three methods are currently defined on these objects:

    $meta->Name;
    $meta->ID;
    $meta->Type;
Name returns the name of the meta as defined in the MetaNames config option when the index was created. The ID is the internal ID number used to represent the meta name. Type is the type of metaname. Currently only one type exists and its value is zero. Swish-e can store content or properties in the index and return this data when running a query. A document's path, URL, title, size, date or summary are examples of properites. Each property is accessed via its PropertyName. This method returns information about the PropertNames stored in the index. Pass in the name of an open index file name and returns a list of SWISH::API::MetaName objects. Three methods are currently defined on these objects:
    $prop->Name;
    $prop->ID;
    $prop->Type;
Name returns the name of the meta as defined in the MetaNames config option when the index was created. The ID is the internal ID number used to represent the meta name. Type is the type of metaname. Currently only one type exists and its value is zero. These also return a list of Property or Metaname description objects, but are accessed via a result record. Since the result comes from a specific index file there's no need to specify the index file name. *Deprecated* Returns the stemmed version of the passed in word. Deprecated because only stems using the original Porter Stemmer and uses a shared memory location in the SW_HANDLE object to store the stemmed word. See below for other stemming options. Like StemWord used to work, only it uses whatever stemmer is named in CW$indexname. Returns the same kind of fuzzy_word object as the FuzzyWord() method. Returns the string (e.g. Stemming_en, Soundex, None ) indicating the stemming method used while indexing the given document. Converts CW$word using the same fuzzy mode used to index the CW$result. Returns a SWISH::API::FuzzyWord object. Methods on the object are used to access the converted words and other data as shown below. Returns the number of output words. Normally this is the value one, but may be more depending on the stemmer used. DoubleMetaphone can return two strings for a single input string. Returns any error code that the stemmer might set. Normally, this return value is zero, indicating that the stemming/fuzzy operation succedded. The values returned are defined in the swish-e source file /src/stemmer.h. Returns the converted words from the stemming/fuzzy operation. Normally, the array will contain a single element, although may contain more (i.e. if DoubleMetaphone is used and the input word returns two strings). In the event that a word does not stem (e.g. trying to stem a number), this method will return the original input word specified when CW$result->FuzzyWord( CW$word ) was called. * Not implemented * Splits up the input string into tokens of swish words and operators.

NOTES

Perl's garbage collection makes it easy to write code for searching with Swish-e, but care must be taken not to keep objects around too long which can use up memory. Here's an example of a potential problem. Say you have a very large number of documents indexed and you want to find the first hit for a number of popular keywords (error checking omitted in this bad example):

    sub first_hit {
      my $query = shift;
      my $handle = SWISH::API->new( 'index.swish-e');
      my $results = $handle->Query( $query );
      my $first_hit = $results->NextResult;
      return $first_hit;
    }
    my @first_hit_list;
    for ( @keywords )
        push @first_hit_list, $first_hit($_);
    }
The first_hit() subroutine is returning a SWISH::Result object. That makes it easy to access properties:
   # print file names
   for my $result ( @first_hit_list ) {
      print $result->Property('swishdocpath'),"\n";
   }
But as long as a SWISH::API::Result object is around, so is the entire list of results generated by the CW$handle->Query() call, and the index file is still open (because a SWISH::API::Result depends on a SWISH::API::Results object, which depends on a SWISH::API object). In this case it would be better to return from first_hit() just the properties you need:
      ...
      my $first_hit = $results->NextResult;
      return $first_hit->Property('swishdocpath');
   }
Then when first_hit() sub ends the result list will be freed, and the index file closed, thanks to Perl's reference count tracking. Note: the other problem with the above code is that the same index file is opened for each call to the function. Don't do that, instead open the index file once.

COPYRIGHT

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

AUTHOR

Bill Moseley moseley@hank.org. 2002/2003/2004

SUPPORT

Please contact the Swish-e discussion email list for support with this module or with Swish-e. Please do not contact the developers directly.