More notes about query model.

[idzebra-moved-to-github.git] / doc / querymodel.xml

 <chapter id="querymodel">
  <!-- $Id: querymodel.xml,v 1.4 2006-06-14 13:44:15 adam Exp $ -->
  <title>Query Model</title>
  
  <sect1 id="querymodel-overview">
   <title>Query Model Overview</title>
   
   <para>
    Zebra is born as a networking Information Retrieval engine adhering
    to the international standards 
    <ulink url="&url.z39.50;">Z39.50</ulink> and
    <ulink url="&url.sru;">SRU</ulink>,
    and implement the query model defined there.
    Unfortunately, the Z39.50 query model has only defined a binary
    encoded representation, which is used as transport packaging in
    the Z39.50 protocol layer. This representation is not human
    readable, nor defines any convenient way to specify queries. 
   </para>
   <!-- tell about RPN - include link to YAZ 
        url.yaz.pqf -->
   <para>
    Therefore, Index Data has defined a textual representation of the
    RPN query: <literal>Prefix Query Format</literal>, short
    <literal>PQF</literal>, which then has been adopted by other
    parties developing Z39.50 software. It is also often referred to as
    <literal>Prefix Query Notation</literal>, or in short 
    <literal>PQN</literal>, and is thoroughly explained in       
    <xref linkend="querymodel-pqf"/>. 
   </para>

   <!-- PQF/RPN is natively supported. CQL is NOT . So we need a map -->
   <para>
    In addition, Zebra can be configured to understand and map the 
    <literal>Common Query Language</literal>
    (<ulink url="&url.cql;">CQL</ulink>)
    to PQF. See an introduction on the mapping to the internal query
    representation in  
    <xref linkend="querymodel-cql-to-pqf"/>.
   </para>
  </sect1>
  
  <sect1 id="querymodel-pqf">
   <title>Prefix Query Format structure and syntax</title>
   <para>
    The <ulink url="&url.yaz.pqf;">PQF grammer</ulink>
    is documented in the YAZ manual, and shall not be
    repeated here. This textual PQF representation
    is always during search mapped to the equivalent Zebra internal
    query parse tree. 
   </para>
   
   <sect2 id="querymodel-pqf-tree">
    <title>PQF tree structure</title>
    <para>
     The PQF parse tree - or the equivalent textual representation -
     may start with one specification of the 
     <emphasis>attribute set</emphasis> used. Following is a query
     tree, which 
     consists of <emphasis>atomic query parts</emphasis>, eventually
     paired by <emphasis>boolean binary operators</emphasis>, and 
     finally  <emphasis>recursively combined </emphasis> into 
     complex query trees.   
    </para>
    
    <sect3 id="querymodel-attribute-sets">
     <title>Attribute sets</title>
     <para>
      Attribute sets define the exact meaning and semantics of queries
      issued. Zebra comes with some predefined attribute set
      definitions, others can easily be defined and added to the
      configuration.
      <note>
       The Zebra internal query procesing is modeled after 
       the <literal>Bib1</literal> attribute set, and the non-use
       attributes type 2-9 are hard-wired in. It is therefore essential
       to be familiar with <xref linkend="querymodel-bib1"/>. 
      </note>
     </para>
     
     <table id="querymodel-attribute-sets-table">
      <caption>Attribute sets predefined in Zebra</caption>
       <!--
       <thead>
       <tr><td>one</td><td>two</td></tr>
      </thead>
       -->
       <tbody>
        <tr>
         <td><emphasis>exp-1</emphasis></td>
         <td><literal>Explain</literal> attribute set</td>
         <td>Special attribute set used on the special automagic
          <literal>IR-Explain-1</literal> database to gain information on
          server capabilities, database names, and database
          and semantics.</td>
        </tr>
        <tr>
         <td><emphasis>bib-1</emphasis></td>
         <td><literal>Bib1</literal> attribute set</td>
         <td>Standard PQF query language attribute set which defines the
          semantics of Z39.50 searching. In addition, all of the
          non-use attributes (type 2-9) define the Zebra internal query
          processing</td>
        </tr>
        <tr>
         <td><emphasis>gils</emphasis></td>
         <td><literal>GILS</literal> attribute set</td>
         <td>Extention to the <literal>Bib1</literal> attribute set.</td>
        </tr>
       </tbody>
     </table>
    </sect3>
    
    <sect3 id="querymodel-boolean-operators">
     <title>Boolean operators</title>
     <para>
      A pair of subquery trees, or of atomic queries, is combined
      using the standard boolean operators into new query trees.
     </para>
     
     <table id="querymodel-boolean-operators-table">
      <caption>Boolean operators</caption>
       <!--
       <thead>
       <tr><td>one</td><td>two</td></tr>
      </thead>
       -->
       <tbody>
        <tr><td><emphasis>@and</emphasis></td>
         <td>binary <literal>AND</literal> operator</td>
         <td>Set intersection of two atomic queries hit sets</td>
        </tr>
        <tr><td><emphasis>@or</emphasis></td>
         <td>binary <literal>OR</literal> operator</td>
         <td>Set union of two atomic queries hit sets</td>
        </tr>
        <tr><td><emphasis>@not</emphasis></td>
         <td>binary <literal>AND NOT</literal> operator</td>
         <td>Set complement of two atomic queries hit sets</td>
        </tr>
        <tr><td><emphasis>@prox</emphasis></td>
         <td>binary <literal>PROXIMY</literal> operator</td>
         <td>Set intersection of two atomic queries hit sets. In 
          addition, the intersection set is purged for all 
          documents which do not satisfy the requested query 
          term proximity. Usually a proper subset of the AND 
          operation.</td>
        </tr>
       </tbody>
     </table>
     
     <para>
      For example, we can combine the terms 
      <emphasis>information</emphasis> and <emphasis>retrieval</emphasis> 
      into different searches in the default index of the default
      attribute set as follows.
      Querying for the union of all documents containing the
      terms <emphasis>information</emphasis> OR
      <emphasis>retrieval</emphasis>: 
      <screen>
       Z> find @or information retrieval
      </screen>
     </para>
     <para>
      Querying for the intersection of all documents containing the
      terms <emphasis>information</emphasis> AND
      <emphasis>retrieval</emphasis>: 
      The hit set is a subset of the coresponding
      OR query.
      <screen>
       Z> find @and information retrieval
      </screen>
     </para>
     <para>
      Querying for the intersection of all documents containing the
      terms <emphasis>information</emphasis> AND
      <emphasis>retrieval</emphasis>, taking proximity into account:
      The hit set is a subset of the coresponding
      AND query.
      <screen>
       Z> find @prox information retrieval
      </screen>
     </para>
     <para>
      Querying for the intersection of all documents containing the
      terms <emphasis>information</emphasis> AND
      <emphasis>retrieval</emphasis>, in the same order and near each
      other as described in the term list  
      The hit set is a subset of the coresponding
      PROXIMY query.
      <screen>
       Z> find "information retrieval"
      </screen>
     </para>
    </sect3>
    
    
    <sect3 id="querymodel-atomic-queries">
     <title>Atomic queries</title>
     <para>
      Atomic queries are the query parts which work on one acess point
      only. These consist of <literal>an attribute list</literal>
      followed by a <literal>single term</literal> or a
      <literal>quoted term list</literal>.
     </para>
     <para>
      Unsupplied non-use attributes type 2-9 are either inherited from
      higher nodes in the query tree, or are set to Zebra's default values.
      See <xref linkend="querymodel-bib1"/> for details. 
     </para>
     
     <table id="querymodel-atomic-queries-table">
      <caption>Atomic queries</caption>
       <!--
       <thead>
       <tr><td>one</td><td>two</td></tr>
      </thead>
       -->
       <tbody>
        <tr><td><emphasis>attribute list</emphasis></td>
         <td>List of <literal>orthogonal</literal> attributes</td>
         <td>Any of the orthogonal attribute types may be omitted,
          these are inherited from higher query tree nodes, or if not
          inherited, are set to the default Zebra configuration values.
         </td>
        </tr>
        <tr><td><emphasis>term</emphasis></td>
         <td>single <literal>term</literal> 
          or <literal>quoted term list</literal>   </td>
         <td>Here the search terms or list of search terms is added
          to the query</td>
        </tr>
       </tbody>
     </table>
     <para>
      Querying for the term <emphasis>information</emphasis> in the
      default index using the default attribite set, the server choice
      of access point/index, and the default non-use attributes.
      <screen>
       Z> find "information"
      </screen>
     </para>
     <para>
      Equivalent query fully specified:
      <screen>
       Z> find @attrset bib-1 @attr 1=1017 @attr 2=3 @attr 3=3 @attr 4=1 @attr 5=100 @attr 6=1 "information"
      </screen>
     </para>
     
     <para>
      Finding all documents which have empty titles. Notice that the
      empty term must be quoted, but is otherwise legal.
      <screen>
       Z> find @attr 1=4 ""
      </screen>
     </para>

    </sect3>
    
    <sect3 id="querymodel-use-string">
     <title>Zebra's special use attribute type 1 of form 'string'</title>
     <para>
      The numeric <literal>use (type 1)</literal> attribute is usually 
      refered to from a given
      attribute set. In addition, Zebra let you use 
      <emphasis>any internal index
       name defined in your configuration</emphasis> 
      as use atribute value. This is a great feature for
      debugging, and when you do
      not need the complecity of defined use attribute values. It is
      the preferred way of accessing Zebra indexes directly.  
     </para>
     <para>
      Finding all documents which have the term list "information
      retrieval" in an Zebra index, using it's internal full string name.
      <screen>
       Z> find @attr 1=sometext "information retrieval"
      </screen>
     </para>
     <para>
      Searching the bib-1 use attribute 54 using it's string name:
      <screen>
       Z> find @attr 1=Code-language eng
      </screen>
     </para>
     <para>
      Searching in any silly string index - if it's defined in your
      indexation rules and can be parsed by the PQF parser. 
      This is definitely not the recommended use of
      this facility, as it might confuse your users with some very
      unexpected results.
      <screen>
       Z> find @attr 1=silly/xpath/alike[@index]/name "information retrieval"
      </screen>
     </para>
     <para>
      See <xref linkend="querymodel-bib1-mapping"/> for details, and 
      <xref linkend="server-sru"/>
      for the SRU PQF query extention using string names as a fast
      debugging facility.
     </para>
    </sect3>
    
    <sect3 id="querymodel-use-xpath">
     <title>Zebra's special use attribute type 1 of form 'XPath' 
      for GRS filters</title>
     <para>
      As we have seen above, it is possible (albeit seldom a great
      idea) to emulate 
      <ulink url="http://www.w3.org/TR/xpath">XPath 1.0</ulink> based
      search by defining <literal>use (type 1)</literal> 
      <emphasis>string</emphasis> attributes which in appearence 
      <emphasis>resemble XPath queries</emphasis>. There are two
      problems with this approach: first, the XPath-look-alike has to
      be defined at indexation time, no new undefined
      XPath queries can entered at search time, and second, it might
      confuse users very much that an XPath-alike index name in fact
      gets populated from a possible entirely different XML element
      than it pretends to acess. 
     </para>
     <para>
      When using the <literal>GRS Record Model</literal> 
      (see  <xref linkend="record-model-grs"/>), we have the
      possibility to embed <emphasis>life</emphasis> 
      XPath expressions
      in the PQF queries, which are here called
      <literal>use (type 1)</literal> <emphasis>xpath</emphasis>
      attributes. You must enable the 
      <literal>xpath enable</literal> directive in your 
      <literal>.abs</literal> config files. 
     </para>
     <note>
      Only a <emphasis>very</emphasis> restricted subset of the 
      <ulink url="http://www.w3.org/TR/xpath">XPath 1.0</ulink> 
      standard is supported as the GRS record model is simpler than
      a full XML DOM structure. See the following examples for 
      possibilities.
     </note>
     <para>
      Finding all documents which have the term "content" 
      inside a text node found in a specific XML DOM
      <emphasis>subtree</emphasis>, whose starting element is 
      adressed by XPath. 
      <screen>
       Z> find @attr 1=/root content 
       Z> find @attr 1=/root/first content
      </screen>
      <emphasis>Notice that the
       XPath must be absolute, i.e., must start with '/', and that the
       XPath <literal>decendant-or-self</literal> axis followed by a
       text node selection <literal>text()</literal> is implicitly
       appended to the stated XPath.
      </emphasis>
      It follows that the above searches are interpreted as:
      <screen>
       Z> find @attr 1=/root//text() content 
       Z> find @attr 1=/root/first//text() content
      </screen>
     </para>
     
     <para>     
      Filter the adressing XPath by a predicate working on exact
      string values in
      attributes (in the XML sense) can be done: return all those docs which
      have the term "english" contained in one of all text subnodes of
      the subtree defined by the XPath
      <literal>/record/title[@lang='en']</literal> 
      <screen>
       Z> find @attr 1=/record/title[@lang='en'] english
      </screen>
     </para>
     
     <para>     
      Combining numeric indexes, boolean expressions, 
      and xpath based searches is possible:
      <screen>
       Z> find @attr 1=/record/title @and foo bar
       Z> find @and @attr 1=/record/title foo @attr 1=4 bar
      </screen> 
     </para>
     <para>
      Escaping PQF keywords and other non-parseable XPath constructs
      with <literal>'{ }'</literal> to prevent syntax errors:
      <screen>
       Z> find @attr {1=/root/first[@attr='danish']} content
       Z> find @attr {1=/root/second[@attr='danish lake']} 
       Z> find @attr {1=/root/third[@attr='dansk s\xc3\xb8']} 
      </screen>
     </para>
     <warning>
      It is worth mentioning that these dynamic performed XPath
      queries are a performance bottelneck, as no optimized
      specialized indexes can be used. Therefore, avoid the use of
      this facility when speed is essential, and the database content
      size is medium to large. 
     </warning>
    </sect3>
    
   </sect2>
   
   <sect2 id="querymodel-exp1">
    <title>Explain Attribute Set</title>
    <para>
     The Z39.50 standard defines the  
     <ulink url="&url.z39.50.explain;">Explain</ulink>attribute set
     <literal>exp-1</literal>, which is used to discover information 
     about a server's search semantics and functional capabilities
     Zebra exposes a  "classic" 
     Explain database by base name <literal>IR-Explain-1</literal>, which
     is populated with system internal information.  
    </para>
   <para>
     The attribute-set <literal>exp-1</literal> consists of a single 
     <literal>Use (type 1)</literal> attribute. 
    </para>
    <para>
     In addition, the non-Use
     <literal>bib-1</literal> attributes, that is, the types 
     <literal>Relation</literal>, <literal>Position</literal>,
     <literal>Structure</literal>, <literal>Truncation</literal>, 
     and <literal>Completeness</literal> are imported from 
     the <literal>bib-1</literal> attribute set, and may be used
     within any explain query. 
    </para>
    
    <sect3 id="querymodel-exp1-use">
    <title>Use Attributes (type = 1)</title>
     <para>
      The following Explain search atributes are supported:
      <literal>ExplainCategory</literal> (@attr 1=1), 
      <literal>DatabaseName</literal> (@attr 1=3), 
      <literal>DateAdded</literal> (@attr 1=9), 
      <literal>DateChanged</literal>(@attr 1=10).
     </para>
     <para>
      A search in the use attribute  <literal>ExplainCategory</literal>
      supports only these predefined values:
      <literal>CategoryList</literal>, <literal>TargetInfo</literal>,
      <literal>DatabaseInfo</literal>, <literal>AttributeDetails</literal>.
     </para>
     <para>
      See <filename>tab/explain.att</filename> and the 
      <ulink url="&url.z39.50;">Z39.50</ulink> standard
      for more information.
     </para>
    </sect3>
    
    <sect3>
     <title>Explain searches with yaz-client</title>
     <para>
      Classic Explain only defines retrieval of Explain information
      via ASN.1. Pratically no Z39.50 clients supports this. Fortunately
      they don't have to - Zebra allows retrieval of this information
      in other formats:
      <literal>SUTRS</literal>, <literal>XML</literal>, 
      <literal>GRS-1</literal> and  <literal>ASN.1</literal> Explain.
     </para>
     
     <para>
      List supported categories to find out which explain commands are
      supported: 
      <screen>
       Z> base IR-Explain-1
       Z> find @attr exp1 1=1 categorylist
       Z> form sutrs
       Z> show 1+2
      </screen>
     </para>
     
     <para>
      Get target info, that is, investigate which databases exist at
      this server endpoint:
      <screen>
       Z> base IR-Explain-1
       Z> find @attr exp1 1=1 targetinfo
       Z> form xml
       Z> show 1+1
       Z> form grs-1
       Z> show 1+1
       Z> form sutrs
       Z> show 1+1
      </screen>
     </para>
     
     <para>
      List all supported databases, the number of hits
      is the number of databases found, which most commonly are the
      following two:
      the <literal>Default</literal> and the
      <literal>IR-Explain-1</literal> databases.
      <screen>
       Z> base IR-Explain-1
       Z> find @attr exp1 1=1 databaseinfo
       Z> form sutrs
       Z> show 1+2
      </screen>
     </para>
     
     <para>
      Get database info record for database <literal>Default</literal>.
      <screen>
       Z> base IR-Explain-1
       Z> find @and @attr exp1 1=1 databaseinfo @attr exp1 1=3 Default
      </screen>
      Identical query with explicitly specified attribute set:
      <screen>
       Z> base IR-Explain-1
       Z> find @attrset exp1 @and @attr 1=1 databaseinfo @attr 1=3 Default
      </screen>
     </para>
     
     <para>
      Get attribute details record for database
      <literal>Default</literal>.
      This query is very useful to study the internal Zebra indexes.
      If records have been indexed using the <literal>alvis</literal>
      XSLT filter, the string representation names of the known indexes can be
      found.
      <screen>
       Z> base IR-Explain-1
       Z> find @and @attr exp1 1=1 attributedetails @attr exp1 1=3 Default
      </screen>
      Identical query with explicitly specified attribute set:
      <screen>
       Z> base IR-Explain-1
       Z> find @attrset exp1 @and @attr 1=1 attributedetails @attr 1=3 Default
      </screen>
     </para>
    </sect3>
    
   </sect2>
   
   <sect2 id="querymodel-bib1">
    <title>Bib1 Attribute Set</title>
    <para>
     Something about querying to be written ..
    </para>
    <para>
     Most of the information contained in this section is an excerpt of
     the <literal>ATTRIBUTE SET BIB-1 (Z39.50-1995)
      SEMANTICS</literal>, 
     found at  <ulink url="&url.z39.50.attset.bib1.1995;">. The BIB-1
      Attribute Set Semantics</ulink> from 1995, also in an updated 
     <ulink url="&url.z39.50.attset.bib1;">Bib-1
      Attribute Set</ulink> 
     version from 2003. Index Data is not the copyright holder of this
     information. 
    </para>
    
    
   <sect3 id="querymodel-bib1-use">
     <title>Use Attributes (type = 1)</title>
    </sect3>

    <para>
     Phrase search for <emphasis>information retrieval</emphasis> in
     the title-register:
     <screen>
      Z> find @attr 1=4 "information retrieval"
     </screen>
    </para>

    
    <sect3 id="querymodel-bib1-relation">
     <title>Relation Attributes (type = 2)</title>
    </sect3>
    <para>
     Supported operations: = (default, of omitted), &lt; &gt; &lt;=, &gt;= .
     Unsupported: Not equal.
     
     The following relation attributes are also supported: relevance (102).
     <!-- always-matches (103) not supported for all indexes -->

     All operations are based on a lexicographical ordering, 
     <emphasis>expect</emphasis> in the case for the
     following structure attributes: numeric(109).

    
    </para>
    
    <para>
     Ranked search for <emphasis>information retrieval</emphasis> in
     the title-register
     (see <xref linkend="administration-ranking"/> for the glory details):
     <screen>
      Z> find @attr 1=4 @attr 2=102 "information retrieval"
     </screen>
    </para>
    
    <sect3 id="querymodel-bib1-position">
     <title>Position Attributes (type = 3)</title>
     <para>
      Only value of (any position(3) is supported. first in field(1),
      and first in subfield(2) are unsupported but using them
      does not trigger an error.
      <!-- It should -->
    </sect3>
    
    <sect3 id="querymodel-bib1-structure">
     <title>Structure Attributes (type = 4)</title>
     <!-- See tab/default.idx -->
    </sect3>
    
    <para>
     For example, in
     the GILS schema (<literal>gils.abs</literal>), the
     west-bounding-coordinate is indexed as type <literal>n</literal>,
     and is therefore searched by specifying
     <emphasis>structure</emphasis>=<emphasis>Numeric String</emphasis>.
     To match all those records with west-bounding-coordinate greater
     than -114 we use the following query:
     <screen>
      Z> find @attr 4=109 @attr 2=5 @attr gils 1=2038 -114
     </screen> 
    </para>

    <sect3 id="querymodel-bib1-truncation">
     <title>Truncation Attributes (type = 5)</title>
     <para>
      Supported are: No truncation(100) which is the default,
      Right trunation(1), Left truncation(2),
      Left&amp;Right truncation(3), 
      Process <literal>#</literal> in term(100) which maps
      each # to <literal>.*</literal>,
      Regexp-1(102) normal regular, Regexp-2(103) (regular with fuzzy),
      <!--
      Special 104, 105, 106 are deprecated and will be removed! -->
      
    </sect3>
    
    <sect3 id="querymodel-bib1-completeness">
    <title>Completeness Attributes (type = 6)</title>
     <para>
      This attribute is ONLY used if structure w, p is to be
      chosen. completeness is ignorned if not w, p is to be
      used..
      Incomplete field(1) is the default and makes Zebra use
      register type w.
      complete subfield(2) and complete field(3) both triggers
      search field type p.
    </sect3>
   </sect2>
    

   <sect2 id="querymodel-zebra-attr-search">
    <title>Zebra specific Search Extentions to all Attribute Sets</title>
    <para>
     Zebra extends the Bib1 attribute types, and these extentions are
     recognized regardless of attribute 
     set used in a <literal>search</literal> operation query.
    </para>

     <table id="querymodel-zebra-attr-search-table">
      <caption>Zebra Search Attribute Extentions</caption>
       <thead>
        <tr>
         <td><emphasis>Name and Type</emphasis></td>
         <td>Operation</td>
         <td>Zebra version</td>
        </tr>
      </thead>
       <tbody>
        <tr>
         <td><emphasis>Embedded Sort (type 7)</emphasis></td>
         <td>search</td>
         <td>1.1</td>
        </tr>
        <tr>
         <td><emphasis>Term Set (type 8)</emphasis></td>
         <td>search</td>
         <td>1.1</td>
        </tr>
        <tr>
         <td><emphasis>Rank weight  (type 9)</emphasis></td>
         <td>search</td>
         <td>1.1</td>
        </tr>
        <tr>
         <td><emphasis>Approx Limit (type 9)</emphasis></td>
         <td>search</td>
         <td>1.4</td>
        </tr>
        <tr>
         <td><emphasis>Term Reference (type 10)</emphasis></td>
         <td>search</td>
         <td>1.4</td>
        </tr>
       </tbody>
      </table>      

    <sect3 id="querymodel-zebra-attr-sorting">
     <title>Zebra Extention Embedded Sort Attribute (type 7)</title>
    </sect3>
    <para>
     The embedded sort is a way to specify sort within a query - thus
     removing the need to send a Sort Request separately. It is both
     faster and does not require clients to deal with the Sort
     Facility. 
    </para>
    <para>
     The possible values after attribute <literal>type 7</literal> are
     <literal>1</literal> ascending and 
     <literal>2</literal> descending. 
     The attributes+term (APT) node is separate from the
     rest and must be <literal>@or</literal>'ed. 
     The term associated with APT is the sorting level in integers,
     where <literal>0</literal> means primary sort, 
     <literal>1</literal> means secondary sort, and so forth.
     See also <xref linkend="administration-ranking"/>.
    </para>
    <para>
     For example, searching for water, sort by title (ascending) 
     <screen>
      Z> find @or @attr 1=1016 water @attr 7=1 @attr 1=4 0
     </screen>
    </para>
    <para>
     Or, searching for water, sort by title ascending, then date descending
     <screen>
      Z> find @or @or @attr 1=1016 water @attr 7=1 @attr 1=4 0 @attr 7=2 @attr 1=30 1
     </screen>
    </para>
    
    <sect3 id="querymodel-zebra-attr-estimation">
     <title>Zebra Extention Term Set Attribute (type 8)</title>
    </sect3>
    <para>
     The Term Set feature is a facility that allows a search to store
     hitting terms in a "pseudo" resultset; thus a search (as usual) +
     a scan-like facility. Requires a client that can do named result
     sets since the search generates two result sets. The value for
     attribute 8 is the name of a result set (string). The terms in
     the named term set are returned as SUTRS records. 
    </para>
    <para>
     For example, searching  for u in title, right truncated, and
     storing the result in term set named 'aset'
     <screen> 
      Z> find @attr 5=1 @attr 1=4 @attr 8=aset u
     </screen>
    </para>
    <warning>
     The model has one serious flaw: we don't know the size of term
     set. Experimental. Do not use in production code.
    </warning>

    <sect3 id="querymodel-zebra-attr-weight">
     <title>Zebra Extention Rank Weight Attribute (type 9)</title>
    </sect3>
    <para>
     Rank weight is a way to pass a value to a ranking algorithm - so
     that one APT has one value - while another as a different one. 
     See also <xref linkend="administration-ranking"/>.
    </para>
    <para>
     For example, searching  for utah in title with weight 30 as well
     as any with weight 20: 
     <screen>  
      Z> find @attr 2=102 @or @attr 9=30 @attr 1=4 utah @attr 9=20 utah
     </screen>
    </para>

    <sect3 id="querymodel-zebra-attr-limit">
     <title>Zebra Extention Approximative Limit Attribute (type 9)</title>
    </sect3>
    <para>
     Newer Zebra versions normally estemiates hit count for every APT
     (leaf) in the query tree. These hit counts are returned as part of
     the searchResult-1 facility in the binary encoded Z39.50 search
     response packages.
    </para>
    <para>
     By setting a limit for the APT we can make Zebra turn into
     approximate hit count when a certain hit count limit is
     reached. A value of zero means exact hit count.
    </para>
    <para>
     For example, we might be intersted in exact hit count for a, but
     for b we allow hit count estimates for 1000 and higher. 
     <screen>
      Z> find @and a @attr 9=1000 b
     </screen>
    </para>
    <note>
     The estimated hit count fascility makes searches faster, as one
     only needs to process large hit lists partially.
    </note>
    <warning>
     This facility clashes with rank weight, because there all
     documents in the hit lists need to be examined for scoring and
     re-sorting.
     It is an experimental
     extention. Do not use in production code.
    </warning>

    <sect3 id="querymodel-zebra-attr-termref">
     <title>Zebra Extention Term Reference Attribute (type 10)</title>
    </sect3>
    <para>
     Zebra supports the searchResult-1 facility. If attribute 10 is
     given, that specifies a subqueryId value returned as part of the
     search result. It is a way for a client to name an APT part of a
     query. 
    </para>
    <!--
    <para>
     <screen>
     </screen>
    </para>
    -->
    <warning>
     Experimental. Do not use in production code.
    </warning>


   </sect2>
    

   <sect2 id="querymodel-zebra-attr-scan">
    <title>Zebra specific Scan Extentions to all Attribute Sets</title>
    <para>
     Zebra extends the Bib1 attribute types, and these extentions are
     recognized regardless of attribute 
     set used in a <literal>scan</literal> operation query.
    </para>
     <table id="querymodel-zebra-attr-scan-table">
      <caption>Zebra Scan Attribute Extentions</caption>
       <thead>
        <tr>
         <td><emphasis>Name and Type</emphasis></td>
         <td>Operation</td>
         <td>Zebra version</td>
        </tr>
      </thead>
       <tbody>
        <tr>
         <td><emphasis>Result Set Narrow (type 8)</emphasis></td>
         <td>scan</td>
         <td>1.3</td>
        </tr>
        <tr>
         <td><emphasis>Approximative Limit (type 9)</emphasis></td>
         <td>scan</td>
         <td>1.4</td>
        </tr>
       </tbody>
      </table>      

    <sect3 id="querymodel-zebra-attr-xyz">
     <title>Zebra Extention Result Set Narrow (type 8)</title>
    </sect3>
    <para>
     If attribute 8 is given for scan, the value is the name of a
     result set. Each hit count in scan is @and'ed with the result set
     given. 
    </para>
    <!--
    <para>
     <screen>
     </screen>
    </para>
    -->
    <warning>
     Experimental and buggy. Definitely not to be used in production code.
    </warning>

    <sect3 id="querymodel-zebra-attr-xyz">
     <title>Zebra Extention Approximative Limit (type 9)</title>
    </sect3>
    <para>
     The approximative limit (as for search) is a way to enable approx
     hit counts for scan hit counts. 
    </para>
    <!--
    <para>
     <screen>
     </screen>
    </para>
    -->
    <warning>
     Experimental. Do not use in production code.
    </warning>


   </sect2>
    

   <sect2 id="querymodel-bib1-mapping">
    <title>Mapping from Bib1 Attributes to Zebra internal 
     register indexes</title>
    <para>
     TO-DO
     </para>


     <!-- see in util/zebramap.c
      int zebra_maps_attr

  if (completeness_value == 2 || completeness_value == 3)
        *complete_flag = 1;
    else
        *complete_flag = 0;
    *reg_id = 0;

    *sort_flag =(sort_relation_value > 0) ? 1 : 0;
    *search_type = "phrase";
    strcpy(rank_type, "void");
    if (relation_value == 102)
    {
        if (weight_value == -1)
            weight_value = 34;
        sprintf(rank_type, "rank,w=%d,u=%d", weight_value, use_value);
    }
    if (relation_value == 103)
    {
        *search_type = "always";
        *reg_id = 'w';
        return 0;
    }
    if (*complete_flag)
        *reg_id = 'p';
    else
        *reg_id = 'w';
    switch (structure_value)
    {
    case 6:   /* word list */
        *search_type = "and-list";
        break;
    case 105: /* free-form-text */
        *search_type = "or-list";
        break;
    case 106: /* document-text */
        *search_type = "or-list";
        break;  
    case -1:
    case 1:   /* phrase */
    case 2:   /* word */
    case 108: /* string */ 
        *search_type = "phrase";
        break;
   case 107: /* local-number */
        *search_type = "local";
        *reg_id = 0;
        break;
    case 109: /* numeric string */
        *reg_id = 'n';
        *search_type = "numeric";
        break;
    case 104: /* urx */
        *reg_id = 'u';
        *search_type = "phrase";
        break;
    case 3:   /* key */
        *reg_id = '0';
        *search_type = "phrase";
        break;
    case 4:  /* year */
        *reg_id = 'y';
        *search_type = "phrase";
        break;
    case 5:  /* date */
        *reg_id = 'd';
        *search_type = "phrase";
        break;
    default:
        return -1;
    }
    return 0;

     -->

    
    <para>
     <emphasis>Use</emphasis> attributes are interpreted according to the
     attribute sets which have been loaded in the
    <literal>zebra.cfg</literal> file, and are matched against specific
     fields as specified in the <literal>.abs</literal> file which
     describes the profile of the records which have been loaded.
     If no Use attribute is provided, a default of Bib-1 Any is assumed.
    </para>
    
    <para>
     If a <emphasis>Structure</emphasis> attribute of
     <emphasis>Phrase</emphasis> is used in conjunction with a
     <emphasis>Completeness</emphasis> attribute of
     <emphasis>Complete (Sub)field</emphasis>, the term is matched
     against the contents of the phrase (long word) register, if one
     exists for the given <emphasis>Use</emphasis> attribute.
     A phrase register is created for those fields in the
     <literal>.abs</literal> file that contains a
     <literal>p</literal>-specifier.
     <!-- ### whatever the hell _that_ is -->
    </para>

    <para>
     If <emphasis>Structure</emphasis>=<emphasis>Phrase</emphasis> is
     used in conjunction with <emphasis>Incomplete Field</emphasis> - the
     default value for <emphasis>Completeness</emphasis>, the
     search is directed against the normal word registers, but if the term
     contains multiple words, the term will only match if all of the words
     are found immediately adjacent, and in the given order.
     The word search is performed on those fields that are indexed as
     type <literal>w</literal> in the <literal>.abs</literal> file.
    </para>

    <para>
     If the <emphasis>Structure</emphasis> attribute is
     <emphasis>Word List</emphasis>,
     <emphasis>Free-form Text</emphasis>, or
     <emphasis>Document Text</emphasis>, the term is treated as a
     natural-language, relevance-ranked query.
     This search type uses the word register, i.e. those fields
     that are indexed as type <literal>w</literal> in the
     <literal>.abs</literal> file.
    </para>

    <para>
     If the <emphasis>Structure</emphasis> attribute is
     <emphasis>Numeric String</emphasis> the term is treated as an integer.
     The search is performed on those fields that are indexed
     as type <literal>n</literal> in the <literal>.abs</literal> file.
    </para>

    <para>
     If the <emphasis>Structure</emphasis> attribute is
     <emphasis>URx</emphasis> the term is treated as a URX (URL) entity.
     The search is performed on those fields that are indexed as type
     <literal>u</literal> in the <literal>.abs</literal> file.
    </para>

    <para>
     If the <emphasis>Structure</emphasis> attribute is
     <emphasis>Local Number</emphasis> the term is treated as
     native Zebra Record Identifier.
    </para>

    <para>
     If the <emphasis>Relation</emphasis> attribute is
     <emphasis>Equals</emphasis> (default), the term is matched
     in a normal fashion (modulo truncation and processing of
     individual words, if required).
     If <emphasis>Relation</emphasis> is <emphasis>Less Than</emphasis>,
     <emphasis>Less Than or Equal</emphasis>,
     <emphasis>Greater than</emphasis>, or <emphasis>Greater than or
      Equal</emphasis>, the term is assumed to be numerical, and a
     standard regular expression is constructed to match the given
     expression.
     If <emphasis>Relation</emphasis> is <emphasis>Relevance</emphasis>,
     the standard natural-language query processor is invoked.
    </para>

    <para>
     For the <emphasis>Truncation</emphasis> attribute,
     <emphasis>No Truncation</emphasis> is the default.
     <emphasis>Left Truncation</emphasis> is not supported.
     <emphasis>Process # in search term</emphasis> is supported, as is
     <emphasis>Regxp-1</emphasis>.
     <emphasis>Regxp-2</emphasis> enables the fault-tolerant (fuzzy)
     search. As a default, a single error (deletion, insertion, 
     replacement) is accepted when terms are matched against the register
     contents.
    </para>
   </sect2>

   <sect2  id="querymodel-regular">
    <title>Zebra Regular Expressions in Truncation Attribute (type = 5)</title>
    
    <para>
     Each term in a query is interpreted as a regular expression if
     the truncation value is either <emphasis>Regxp-1 (@attr 5=102)</emphasis>
     or <emphasis>Regxp-2 (@attr 5=103)</emphasis>.
     Both query types follow the same syntax with the operands:
    </para>

     <table id="querymodel-regular-operands-table">
      <caption>Regular Expression Operands</caption>
       <!--
       <thead>
       <tr><td>one</td><td>two</td></tr>
      </thead>
       -->
       <tbody>
        <tr>
         <td><emphasis>x</emphasis></td>
         <td>Matches the character <emphasis>x</emphasis>.</td>
        </tr>
        <tr>
         <td><emphasis>.</emphasis></td>
         <td>Matches any character.</td>
        </tr>
        <tr>
         <td><emphasis>[ .. ]</emphasis></td>
         <td>Matches the set of characters specified;
         such as <literal>[abc]</literal> or <literal>[a-c]</literal>.</td>
        </tr>
       </tbody>
      </table>      

    <para>
     The above operands can be combined with the following operators:
    </para>

    
     <table id="querymodel-regular-operators-table">
      <caption>Regular Expression Operators</caption>
       <!--
       <thead>
       <tr><td>one</td><td>two</td></tr>
      </thead>
       -->
       <tbody>
        <tr>
         <td><emphasis>x*</emphasis></td>
         <td>Matches <emphasis>x</emphasis> zero or more times. 
          Priority: high.</td>
        </tr>
        <tr>
         <td><emphasis>x+</emphasis></td>
         <td>Matches <emphasis>x</emphasis> one or more times. 
          Priority: high.</td>
        </tr>
        <tr>
         <td><emphasis>x?</emphasis></td>
         <td> Matches <emphasis>x</emphasis> zero or once. 
          Priority: high.</td>
        </tr>
        <tr>
         <td><emphasis>xy</emphasis></td>
         <td> Matches <emphasis>x</emphasis>, then <emphasis>y</emphasis>.
         Priority: medium.</td>
        </tr>
        <tr>
         <td><emphasis>x|y</emphasis></td>
         <td> Matches either <emphasis>x</emphasis> or <emphasis>y</emphasis>.
         Priority: low.</td>
        </tr>
        <tr>
         <td><emphasis>( )</emphasis></td>
         <td>The order of evaluation may be changed by using parentheses.</td>
        </tr>
       </tbody>
      </table>      
    
    <para>
     If the first character of the <emphasis>Regxp-2</emphasis> query
     is a plus character (<literal>+</literal>) it marks the
     beginning of a section with non-standard specifiers.
     The next plus character marks the end of the section.
     Currently Zebra only supports one specifier, the error tolerance,
     which consists one digit. 
    </para>

    <para>
     Since the plus operator is normally a suffix operator the addition to
     the query syntax doesn't violate the syntax for standard regular
     expressions.
    </para>

    <para>
     For example, a phrase search with regular expressions  in
     the title-register is performed like this:
     <screen>
      Z> find @attr 1=4 @attr 5=102 "informat.* retrieval"
     </screen>
    </para>

    <para>
     Combinations with other attributes are possible. For example, a
     ranked search with a regular expression 
     (see <xref linkend="administration-ranking"/> for the glory details):
     <screen>
      Z> find @attr 1=4 @attr 5=102 @attr 2=102 "informat.* retrieval"
     </screen>
    </para>
   </sect2>

  
   <!--
   <para>
    The RecordType parameter in the <literal>zebra.cfg</literal> file, or
    the <literal>-t</literal> option to the indexer tells Zebra how to
    process input records.
    Two basic types of processing are available - raw text and structured
    data. Raw text is just that, and it is selected by providing the
    argument <emphasis>text</emphasis> to Zebra. Structured records are
    all handled internally using the basic mechanisms described in the
    subsequent sections.
    Zebra can read structured records in many different formats.
   </para>
   -->
  </sect1>


  <sect1 id="querymodel-cql-to-pqf">
   <title>Server Side CQL to PQF Query Translation</title>
   <para>
    Using the
    <literal>&lt;cql2rpn&gt;l2rpn.txt&lt;/cql2rpn&gt;</literal>
      YAZ Frontend Virtual
    Hosts option, one can configure
    the YAZ Frontend CQL-to-PQF
    converter, specifying the interpretation of various 
    <ulink url="&url.cql;">CQL</ulink>
    indexes, relations, etc. in terms of Type-1 query attributes.
    <!-- The  yaz-client config file -->  
   </para>
   <para>
    For example, using server-side CQL-to-PQF conversion, one might
    query a zebra server like this:
    <screen>
    <![CDATA[
     yaz-client localhost:9999
     Z> querytype cql
     Z> find text=(plant and soil)
     ]]>
    </screen>
     and - if properly configured - even static relevance ranking can
     be performed using CQL query syntax:
    <screen>
    <![CDATA[
     Z> find text = /relevant (plant and soil)
     ]]>
     </screen>
   </para>

   <para>
    By the way, the same configuration can be used to 
    search using client-side CQL-to-PQF conversion:
    (the only difference is <literal>querytype cql2rpn</literal> 
    instead of 
    <literal>querytype cql</literal>, and the call specifying a local
    conversion file)
    <screen>
    <![CDATA[
     yaz-client -q local/cql2pqf.txt localhost:9999
     Z> querytype cql2rpn
     Z> find text=(plant and soil)
     ]]>
     </screen>
   </para>

   <para>
    Exhaustive information can be found in the
    Section "Specification of CQL to RPN mappings" in the YAZ manual.
    <ulink url="http://www.indexdata.dk/yaz/doc/tools.tkl#tools.cql.map">
     http://www.indexdata.dk/yaz/doc/tools.tkl#tools.cql.map</ulink>,
   and shall therefore not be repeated here.
   </para> 
  <!-- 
  <para>
    See 
      <ulink url="http://www.loc.gov/z3950/agency/zing/cql/dc-indexes.html">
      http://www.loc.gov/z3950/agency/zing/cql/dc-indexes.html</ulink>
    for the Maintenance Agency's work-in-progress mapping of Dublin Core
    indexes to Attribute Architecture (util, XD and BIB-2)
    attributes.
   </para>
   -->
 </sect1>



</chapter>

 <!-- Keep this comment at the end of the file
 Local variables:
 mode: sgml
 sgml-omittag:t
 sgml-shorttag:t
 sgml-minimize-attributes:nil
 sgml-always-quote-attributes:t
 sgml-indent-step:1
 sgml-indent-data:t
 sgml-parent-document: "zebra.xml"
 sgml-local-catalogs: nil
 sgml-namecase-general:t
 End:
 -->