Github user mktal commented on a diff in the pull request:
https://github.com/apache/incubator-madlib/pull/29#discussion_r57405726
--- Diff: src/ports/postgres/modules/utilities/path.py_in ---
@@ -118,140 +120,172 @@ def path(schema_madlib, source_table, output_table,
partition_expr,
# string produced by concatenating the symbols. The exact rows
that
# produce the match are identified by correlating the matched
string
# indices with another array containing row ids.
- #
- # matched_partitions: For each partition (group), concatenate
all symbols
- # into a single string (sym_str). Keep corresponding ids in
an array in the
- # same order as the symbols. This is performed only for
partitions
- # that contain a match.
- # build_multiple_matched_rows:
- # q1: Split sym_str into an array containing the lengths of
the
- # strings between the matches.
- # q2: Store lengths of matches into an array
- # q3: Merge q1 and q2 and unnest the arrays (ensuring same
length).
- # Also right shift the matches array.
- # q4: Compute the cumulative sum of the arrays.
+
+ # matched_partitions: For each partition, concatenate all symbols
+ # into a single string (sym_str). Keep corresponding ids in
an
+ # array (match_to_row_id) in the same order as the symbols.
+ # This is performed only for partitions that contain a match.
+ match_id_name = "__madlib_path_match_id__" if "match_id" in
all_input_cols else "match_id"
+ symbol_name = "__madlib_path_symbol__" if "symbol" in
all_input_cols else "symbol"
plpy.execute("""
CREATE TEMP TABLE {matched_partitions} AS
SELECT
{p_col_name_str},
- array_to_string(array_agg({symbol_name_str} ORDER BY
{order_expr}), '') as sym_str,
- array_agg({id_col_name} ORDER BY {order_expr}) as
matched_ids
+ array_to_string(array_agg({short_sym_name_str} ORDER
BY {order_expr}), '') as sym_str,
+ array_agg({id_col_name} ORDER BY {order_expr}) as
{match_to_row_id}
FROM {input_with_id}
+ WHERE {short_sym_name_str} is NOT NULL
GROUP BY {p_col_name_str}
- HAVING array_to_string(array_agg({symbol_name_str} ORDER
BY {order_expr}), '') ~* '{pattern_expr}'
+ HAVING array_to_string(array_agg({short_sym_name_str}
ORDER BY {order_expr}), '')
+ ~* '{new_pattern_expr}'
+ """.format(**locals()))
+
+ # length_of_matches: For each partition in matched_partitions:
+ # - find all matches and compute the lengths of each match.
+ # - output these lengths in an array (matches), along with
+ # an array of corresponding rank of each match
(match_indices).
+ length_of_matches = unique_string("match_length_view")
+ plpy.execute("""
+ CREATE VIEW {length_of_matches} AS
+ SELECT
+ {p_col_name_str},
+ {match_to_row_id},
+ array_agg(length(matches) ORDER BY match_index) AS matches,
+ array_agg(match_index ORDER BY match_index) AS
match_indices
+ FROM (
+ SELECT
+ {p_col_name_str},
+ {match_to_row_id},
+ generate_series(1, num_matches) AS match_index,
+ (regexp_matches(
+ sym_str, '(?i)({new_pattern_expr})', 'g'))[1]
AS matches
+ FROM (
+ SELECT
+ {p_col_name_str},
+ {match_to_row_id},
+ sym_str,
+ count(matches) AS num_matches
+ FROM (
+ SELECT
+ {p_col_name_str},
+ {match_to_row_id},
+ sym_str,
+ (regexp_matches(
+ sym_str, '(?i)({new_pattern_expr})',
'g'))[1] AS matches
+ FROM
+ {matched_partitions}
+ ) t1
+ GROUP BY {p_col_name_str}, {match_to_row_id}, sym_str
--- End diff --
I believe this can also be simplified to `group by {p_col_name_str}`
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
