======= Example: Adding a Material Type (947$t) facet index =======

First, every database script should perform its work inside a transaction.  This is to protect the consistency of the data and avoid partial implementation of desired changes in the case of an error.

<code sql>
BEGIN;
</code>

Now we need to add the [[documentation:indexing#indexed_field_definitions|indexing definition]]:

<code sql>
INSERT INTO config.metabib_field (field_class, name, xpath, format, label, search_field, facet_field) VALUES (
  'identifier',                                                -- The class to which this definition will belong
  'mattype',                                                   -- The symbolic name of the definition within the class
  $$//marc:datafield[@tag="947"]/marc:subfield[@code="t"]$$,   -- The XPath expression which will extract the desired data from the MARCXML
  'marcxml',                                                   -- The format in which we will process the XPath for this definition
  'Material Type',                                             -- The translatable label used for this definition in user interfaces
  FALSE,                                                       -- Mark this definition as NOT search-oriented
  TRUE                                                         -- Mark this definition as a facet-oriented index
);
</code>

Allowing the default **id** value to be set by the database gives us the ability to use the **CURRVAL()** function for the auto-incrementing sequence that fills in the **id**.  We can use this when populating the normalization map and the index table with data from the MARC record:

<code sql>
INSERT INTO config.metabib_field_index_norm_map (field, norm, pos)
  SELECT  CURRVAL('config.metabib_field_id_seq'::regclass),  -- Returns the id of the indexing definition inserted above
          cin.id,                                            -- The ids of the normalizers we care about
          -3                                                 -- Ordering of normalizer application -- this one first
    FROM  config.index_normalizer AS cin  -- Table defining normalizer functions
    WHERE cin.name = 'Up-case';           -- The normalizers we care about for this field

INSERT INTO config.metabib_field_index_norm_map (field, norm, params, pos)
  SELECT  CURRVAL('config.metabib_field_id_seq'::regclass),  -- Returns the id of the indexing definition inserted above
          cin.id,                                            -- The ids of the normalizers we care about
          '["[",""]',                                        -- Parameters for the Replace function -- replace "[" with the empty string
          -2                                                 -- Ordering of normalizer application -- this one second
    FROM  config.index_normalizer AS cin  -- Table defining normalizer functions
    WHERE cin.name = 'Replace';           -- The normalizers we care about for this field

INSERT INTO config.metabib_field_index_norm_map (field, norm, params, pos)
  SELECT  CURRVAL('config.metabib_field_id_seq'::regclass),  -- Returns the id of the indexing definition inserted above
          cin.id,                                            -- The ids of the normalizers we care about
          '["]",""]',                                        -- Parameters for the Replace function -- replace "]" with the empty string
          -1                                                 -- Ordering of normalizer application -- this one third
    FROM  config.index_normalizer AS cin  -- Table defining normalizer functions
    WHERE cin.name = 'Replace';           -- The normalizers we care about for this field
</code>  

In order to avoid reindexing the entire bibliographic data set for all index definitions when simply adding a new definition, it is best to backfill the index table for the new definition.  This is done by pulling data from the existing, flattened MARC data:

<code sql>
INSERT INTO metabib.facet_entry (field, source, value)
  SELECT  CURRVAL('config.metabib_field_id_seq'::regclass),  -- Returns the id of the indexing definition inserted above
          mfr.record,                                        -- The source MARC record from which the data will be extracted
          UPPERCASE(BTRIM(value,'[]'))                       -- Normalize 947$t data using the same logic as the defined above for initial index population
    FROM  metabib.real_full_rec AS mfr        -- The internal table holding a flattened version of the bibliographic record
    WHERE mfr.tag = '947' AND subfield = 't'; -- Include only subfield t from 947 tags
</code>

If an alias is desired for targeting exactly this field in SRU and Z39.50 searches, that can be added as well:

<code sql>
INSERT INTO config.metabib_search_alias (alias, field_class, field) VALUES (
  'eg.mattype',                                     -- Alias to accept in SRU/CQL queries, and to map in Z39.50 settings
  'identifier',                                     -- The search class to which the definition belongs
  CURRVAL('config.metabib_field_id_seq'::regclass)  -- Returns the id of the indexing definition inserted above
);
</code>

The only thing left to do is have the database commit the transaction wrapping all of these commands.  The final script thus ends up as:

<code sql>
BEGIN;

INSERT INTO config.metabib_field (field_class, name, xpath, format, label, search_field, facet_field) VALUES (
  'identifier',                                                -- The class to which this definition will belong
  'mattype',                                                   -- The symbolic name of the definition within the class
  $$//marc:datafield[@tag="947"]/marc:subfield[@code="t"]$$,   -- The XPath expression which will extract the desired data from the MARCXML
  'marcxml',                                                   -- The format in which we will process the XPath for this definition
  'Material Type',                                             -- The translatable label used for this definition in user interfaces
  FALSE,                                                       -- Mark this definition as NOT search-oriented
  TRUE                                                         -- Mark this definition as a facet-oriented index
);

INSERT INTO config.metabib_field_index_norm_map (field, norm, pos)
  SELECT  CURRVAL('config.metabib_field_id_seq'::regclass),  -- Returns the id of the indexing definition inserted above
          cin.id,                                            -- The ids of the normalizers we care about
          -3                                                 -- Ordering of normalizer application -- this one first
    FROM  config.index_normalizer AS cin  -- Table defining normalizer functions
    WHERE cin.name = 'Up-case';           -- The normalizers we care about for this field

INSERT INTO config.metabib_field_index_norm_map (field, norm, params, pos)
  SELECT  CURRVAL('config.metabib_field_id_seq'::regclass),  -- Returns the id of the indexing definition inserted above
          cin.id,                                            -- The ids of the normalizers we care about
          '["[",""]',                                        -- Parameters for the Replace function -- replace "[" with the empty string
          -2                                                 -- Ordering of normalizer application -- this one second
    FROM  config.index_normalizer AS cin  -- Table defining normalizer functions
    WHERE cin.name = 'Replace';           -- The normalizers we care about for this field

INSERT INTO config.metabib_field_index_norm_map (field, norm, params, pos)
  SELECT  CURRVAL('config.metabib_field_id_seq'::regclass),  -- Returns the id of the indexing definition inserted above
          cin.id,                                            -- The ids of the normalizers we care about
          '["]",""]',                                        -- Parameters for the Replace function -- replace "]" with the empty string
          -1                                                 -- Ordering of normalizer application -- this one third
    FROM  config.index_normalizer AS cin  -- Table defining normalizer functions
    WHERE cin.name = 'Replace';           -- The normalizers we care about for this field

INSERT INTO metabib.facet_entry (field, source, value)
  SELECT  CURRVAL('config.metabib_field_id_seq'::regclass),  -- Returns the id of the indexing definition inserted above
          mfr.record,                                        -- The source MARC record from which the data will be extracted
          UPPERCASE(BTRIM(value,'[]'))                       -- Normalize 947$t data using the same logic as the defined above for initial index population
    FROM  metabib.real_full_rec AS mfr        -- The internal table holding a flattened version of the bibliographic record
    WHERE mfr.tag = '947' AND subfield = 't'; -- Include only subfield t from 947 tags

INSERT INTO config.metabib_search_alias (alias, field_class, field) VALUES (
  'eg.mattype',                                     -- Alias to accept in SRU/CQL queries, and to map in Z39.50 settings
  'identifier',                                     -- The search class to which the definition belongs
  CURRVAL('config.metabib_field_id_seq'::regclass)  -- Returns the id of the indexing definition inserted above
);

COMMIT;
</code>