LP#1379815 Fetch tag data as a table using tag/xpath combo

New function: oils_xpath_tag_to_table()

This function is adapted from oils_xpath_table() with the goal of being
more targeted and simpler to use.

The main issue with oils_xpath_table() is that it relies on peer
UNNEST() functions, and that leads to unexpected behavior whenever the
xpath arguments result in uneven or "gapped" selections.  In the first
type of case, the resulting table includes rows representing the least
common multiple of the underlying xpath selections.  In the second
type, though the xpaths may sometimes return the same number of values,
those values are not correlated except by order in the marc, which does
not account for the real possibility of null values in the set.

Crude Example:

999 $d ABC $e 123
999 $d DEF
999 $d GHI
999 $e 456

We need a table representing subfields 'd' and 'e' of the '999' fields,
so we might try an xpath like:

//*[@tag="999"]/*[@code="d"]|//*[@tag="999"]/*[@code="e"]

We want:

 d  |  e
---------
ABC | 123
DEF |
GHI |
    | 456

but we get:

 d  |  e
---------
ABC | 123
DEF | 456
GHI | 123
ABC | 456
DEF | 123
GHI | 456

This example illustrates both negative behaviors (non-correlated fields
and least-common-multiple row multiplication).

The new method, while internally quite similar, has a different
signature, with the most significant change being a 'tag' argument
which serves as a common base element for the xpaths (now an array
rather than a pipe-delimited string).

Signed-off-by: Dan Wells <dbw2@calvin.edu>
Signed-off-by: Ben Shum <bshum@biblio.org>

diff --git a/Open-ILS/src/sql/Pg/002.functions.config.sql b/Open-ILS/src/sql/Pg/002.functions.config.sql
index 077e9e8..527e89b 100644 (file)
--- a/Open-ILS/src/sql/Pg/002.functions.config.sql
+++ b/Open-ILS/src/sql/Pg/002.functions.config.sql
@@ -185,6 +185,48 @@ SELECT * FROM (
 END;
 $func$ LANGUAGE PLPGSQL IMMUTABLE;
 
+CREATE OR REPLACE FUNCTION oils_xpath_tag_to_table(marc text, tag text, xpaths text[]) RETURNS SETOF record AS $function$
+
+-- This function currently populates columns with the FIRST matching value
+-- of each XPATH.  It would be reasonable to add a 'return_arrays' option
+-- where each column is an array of all matching values for each path, but
+-- that remains as a TODO
+
+DECLARE
+    field RECORD;
+    output RECORD;
+    select_list TEXT[];
+    from_list TEXT[];
+    q TEXT;
+BEGIN
+    -- setup query select
+    FOR i IN 1 .. ARRAY_UPPER(xpaths,1) LOOP
+        IF xpaths[i] = 'null()' THEN
+            select_list := ARRAY_APPEND(select_list, 'NULL::TEXT AS c_' || i );
+        ELSE
+            select_list := ARRAY_APPEND(select_list, '(oils_xpath(' ||
+                quote_literal(
+                    CASE
+                        WHEN xpaths[i] ~ $re$/[^/[]*@[^/]+$$re$ -- attribute
+                            OR xpaths[i] ~ $re$text\(\)$$re$
+                        THEN xpaths[i]
+                        ELSE xpaths[i] || '//text()'
+                    END
+                ) || ', field_marc))[1] AS cl_' || i);
+                -- hardcoded to first value for each path
+        END IF;
+    END LOOP;
+
+    -- run query over tag set
+    q := 'SELECT ' || ARRAY_TO_STRING(select_list, ',')
+        || ' FROM UNNEST(oils_xpath(' || quote_literal('//*[@tag="' || tag
+        || '"]') || ', ' || quote_literal(marc) || ')) AS field_marc;';
+    --RAISE NOTICE '%', q;
+
+    RETURN QUERY EXECUTE q;
+END;
+
+$function$ LANGUAGE PLPGSQL;
 
 CREATE OR REPLACE FUNCTION extract_marc_field ( TEXT, BIGINT, TEXT, TEXT ) RETURNS TEXT AS $$
 DECLARE

diff --git a/Open-ILS/src/sql/Pg/upgrade/XXXX.function.vandelay-oils_xpath_tag_to_table.sql b/Open-ILS/src/sql/Pg/upgrade/XXXX.function.vandelay-oils_xpath_tag_to_table.sql
new file mode 100644 (file)
index 0000000..48e44f8
--- /dev/null
+++ b/Open-ILS/src/sql/Pg/upgrade/XXXX.function.vandelay-oils_xpath_tag_to_table.sql
@@ -0,0 +1,48 @@
+BEGIN;
+
+--SELECT evergreen.upgrade_deps_block_check('XXXX', :eg_version);
+
+CREATE OR REPLACE FUNCTION oils_xpath_tag_to_table(marc text, tag text, xpaths text[]) RETURNS SETOF record AS $function$
+
+-- This function currently populates columns with the FIRST matching value
+-- of each XPATH.  It would be reasonable to add a 'return_arrays' option
+-- where each column is an array of all matching values for each path, but
+-- that remains as a TODO
+
+DECLARE
+    field RECORD;
+    output RECORD;
+    select_list TEXT[];
+    from_list TEXT[];
+    q TEXT;
+BEGIN
+    -- setup query select
+    FOR i IN 1 .. ARRAY_UPPER(xpaths,1) LOOP
+        IF xpaths[i] = 'null()' THEN
+            select_list := ARRAY_APPEND(select_list, 'NULL::TEXT AS c_' || i );
+        ELSE
+            select_list := ARRAY_APPEND(select_list, '(oils_xpath(' ||
+                quote_literal(
+                    CASE
+                        WHEN xpaths[i] ~ $re$/[^/[]*@[^/]+$$re$ -- attribute
+                            OR xpaths[i] ~ $re$text\(\)$$re$
+                        THEN xpaths[i]
+                        ELSE xpaths[i] || '//text()'
+                    END
+                ) || ', field_marc))[1] AS cl_' || i);
+                -- hardcoded to first value for each path
+        END IF;
+    END LOOP;
+
+    -- run query over tag set
+    q := 'SELECT ' || ARRAY_TO_STRING(select_list, ',')
+        || ' FROM UNNEST(oils_xpath(' || quote_literal('//*[@tag="' || tag
+        || '"]') || ', ' || quote_literal(marc) || ')) AS field_marc;';
+    --RAISE NOTICE '%', q;
+
+    RETURN QUERY EXECUTE q;
+END;
+
+$function$ LANGUAGE PLPGSQL;
+
+COMMIT;