cartodb4/cartodb-postgresql
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Workaround PostgreSQL bug possibly triggering infinite loops

Browse Source 
Implements casting of existing system column values during
cartodbfication using ALTER rather than UPDATE.

It's faster for tables with many rows and works around a
longstanding PostgreSQL bug. See
http://www.postgresql.org/message-id/20140530143150.GA11051@localhost

CDB-3059
This commit is contained in:
Sandro Santilli 2014-05-30 19:09:24 +02:00
parent cc9cbdb7a7
commit ce20009fb2
4 changed files with 30 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
NEWS
View File

@ -7,6 +7,7 @@ Bug fixes:
- Ensure cartodb_id uses an associated sequence (#33)
- Fully qualify call to cdb_disable_ddl_hooks from cdb_enable_ddl_hooks
- Fully qualify call to CDB_UserDataSize from quota trigger
- Fix potential infinite loop in CDB_CartodbfyTable
0.1.0 - 2014-05-23

12
scripts-available/CDB_CartodbfyTable.sql
View File

@ -158,7 +158,11 @@ BEGIN
RAISE NOTICE 'Trying to recover data from % column', new_name;
BEGIN
-- Copy existing values to new field
sql := 'UPDATE ' || reloid::text || ' SET cartodb_id = '
-- NOTE: using ALTER is a workaround to a PostgreSQL bug and
-- is also known to be faster for tables with many rows
-- See http://www.postgresql.org/message-id/20140530143150.GA11051@localhost
sql := 'ALTER TABLE ' || reloid::text
|| ' ALTER cartodb_id TYPE int USING '
|| new_name || '::int4';
RAISE DEBUG 'Running %', sql;
EXECUTE sql;
@ -273,7 +277,11 @@ cname, rec2.typname;
RAISE NOTICE 'Trying to recover data from % coumn', new_name;
BEGIN
-- Copy existing values to new field
sql := 'UPDATE ' || reloid::text || ' SET ' || rec.cname || ' = '
-- NOTE: using ALTER is a workaround to a PostgreSQL bug and
-- is also known to be faster for tables with many rows
-- See http://www.postgresql.org/message-id/20140530143150.GA11051@localhost
sql := 'ALTER TABLE ' || reloid::text || ' ALTER ' || rec.cname
|| ' TYPE TIMESTAMPTZ USING '
|| new_name || '::timestamptz';
RAISE DEBUG 'Running %', sql;
EXECUTE sql;

16
test/CDB_CartodbfyTableTest.sql
View File

@ -200,6 +200,22 @@ SELECT CDB_CartodbfyTableCheck('t', 'unsequenced cartodb_id');
select cartodb_id FROM t;
DROP TABLE t;
-- table with existing the_geom and created_at and containing null values
-- Really, a test for surviving an longstanding PostgreSQL bug:
-- http://www.postgresql.org/message-id/20140530143150.GA11051@localhost
CREATE TABLE t (
the_geom geometry(Geometry,4326),
created_at timestamptz,
updated_at timestamptz
);
COPY t (the_geom, created_at, updated_at) FROM stdin;
0106000020E610000001000000010300000001000000050000009EB8244146435BC017B65E062AD343409EB8244146435BC0F51AF6E2708044400B99891683765AC0F51AF6E2708044400B99891683765AC017B65E062AD343409EB8244146435BC017B65E062AD34340 2012-06-06 21:59:08 2013-06-10 20:17:20
0106000020E61000000100000001030000000100000005000000DA7763431A1A5CC0FBCEE869313C3A40DA7763431A1A5CC09C1B8F55BC494440F9F4A9C7993356C09C1B8F55BC494440F9F4A9C7993356C0FBCEE869313C3A40DA7763431A1A5CC0FBCEE869313C3A40 2012-06-06 21:59:08 2013-06-10 20:17:20
\N \N \N
\.
SELECT CDB_CartodbfyTableCheck('t', 'null geom and timestamp values');
DROP TABLE t;
-- TODO: table with existing custom-triggered the_geom
DROP FUNCTION CDB_CartodbfyTableCheck(regclass, text);

3
test/CDB_CartodbfyTableTest_expect
View File

@ -44,5 +44,8 @@ SELECT 1
unsequenced cartodb_id cartodbfied fine
1
DROP TABLE
CREATE TABLE
null geom and timestamp values cartodbfied fine
DROP TABLE
DROP FUNCTION
DROP FUNCTION