Javier Goizueta
00bd302f01
Avoid creating GridCluster overviews for non-point datasets
2016-01-29 16:19:35 +01:00
Javier Goizueta
93d4a6ead0
Restrict SECURITY DEFINER to overview registration
...
And check it is applied only overview tables with valid names
2016-01-28 17:04:06 +01:00
Javier Goizueta
66387c2d44
Make CDB_CreateOverviews a SECURITY DEFINER function
...
So that it can change the permissions of overview tables
(by updating pg_class) even when executed by non-priviledeged users.
2016-01-28 16:32:40 +01:00
Javier Goizueta
b8d50204dd
Avoid aggregation which causes out-of-memory crashes in PostgreSQL
...
The use of multiple string_agg functions, even if applied to groups
of one single record causes out of memory crashes in PG 9.3.4 for
some (large) tables.
2016-01-27 15:24:04 +01:00
Javier Goizueta
474de01757
Change the overview table naming scheme
...
The scheme is changed from table_ovN to _vovw_N_table for
lower collision probability.
Also future naming changes will be easier by using the functions
_CDB_OverviewTableDiscriminator, _CDB_OverviewTableName,
_CDB_IsOverviewTableOf, etc.
2016-01-26 13:20:28 +01:00
Javier Goizueta
1ebaeb76ac
Overload CDB_Overviews to accept either a single table or an array
...
The result of CDB_Overviews has been expandend with a base_table column
2016-01-26 10:39:26 +01:00
Javier Goizueta
7033a8d9ac
Fixes for table names that require quotes or which include a schema name
...
Some cases left unsolved, with FIXME comments
2016-01-13 18:49:27 +01:00
Javier Goizueta
46bc774d38
Fix CDB_Overviews for the case that the table name requires quoting
2016-01-13 18:24:06 +01:00
Javier Goizueta
a9e2d19918
Document CDB_DropOverviews
2016-01-12 16:20:11 +01:00
Javier Goizueta
5b47c51221
Add function to drop overviews
2016-01-12 15:35:50 +01:00
Javier Goizueta
ccdf8de59e
Enhance documentation, update comments
2016-01-12 12:08:41 +01:00
Javier Goizueta
02f386be33
Regard the GridCluster reduction strategy as private
...
The *strategy* parameteriation will be considered an
implementation detail for the time being.
2016-01-12 11:59:42 +01:00
Javier Goizueta
0a066e0126
Change the default overview reduction strategy to GridCluster
2016-01-11 15:35:04 +01:00
Javier Goizueta
913640e2dc
Preserve the column ordering of the base table in overviews
2016-01-11 15:34:10 +01:00
Javier Goizueta
fd7a8cff71
Change regular expression for consistency
2016-01-11 15:27:05 +01:00
Javier Goizueta
6ea63af974
Add function to obtain existing overviews of a table
2016-01-11 14:50:10 +01:00
Javier Goizueta
496f079b1c
Copy dataset privileges to overview tables
2016-01-04 18:06:56 +01:00
Javier Goizueta
4580c9cd5c
Change the cartodb_ids of aggregated overviews
...
Instead of arbitrary ids, the id of one of the aggregated records
is used, so that if it is used in the UI to query point
the information of one of the records grouped at the point will be
retrieved.
2016-01-04 10:33:39 +01:00
Javier Goizueta
08828b8b7d
Set owner of overlays to the owner of the base table
2015-12-29 15:22:12 +01:00
Javier Goizueta
cecba655eb
Preserve aggregated attributes of singleton groups
2015-12-29 14:56:47 +01:00
Javier Goizueta
b34a752172
Generate overviews for all Z levels
...
Skipping levels produces visually inferior results
2015-12-29 14:56:04 +01:00
Javier Goizueta
8ba9e74c4a
Order columns of aggregated overlays as in the base table
2015-12-29 13:38:07 +01:00
Javier Goizueta
6a6a5bc96a
Fix Sampling reduction to avoid RandomTids problems
...
The fixed cases will not be common but do occur in tests.
This is an interim fix which should be reverted if CDB_randomTids changes.
2015-12-28 19:41:14 +01:00
Javier Goizueta
2ff686de27
Clean up: remove spurious comment
2015-12-28 19:30:20 +01:00
Javier Goizueta
07326626b7
More efficient sampling strategy
...
It is also renamed properly.
2015-12-28 19:28:20 +01:00
Javier Goizueta
4afc427008
Rename Ref. Z strategy function
...
It's not that *dummy* anymore, so choose a more descriptive name
2015-12-28 17:47:57 +01:00
Javier Goizueta
a8a2c04d71
Remove invalid comment
2015-12-23 14:20:57 +01:00
Javier Goizueta
a5bca7d715
Add function for post-processing new overview tables
2015-12-23 14:17:50 +01:00
Javier Goizueta
552206464e
Make strategy function public
2015-12-23 12:50:04 +01:00
Javier Goizueta
2af0b9a57f
Add function comments
2015-12-23 12:42:40 +01:00
Javier Goizueta
1b5de84c9a
Add missing attribute-aggregationto the point reduction strategy
2015-12-23 12:17:37 +01:00
Javier Goizueta
5a78ee2896
Optimize the gridded clustering strategy
...
The internal grid_px parameter is adjusted for best results with default symbol size
2015-12-22 17:59:49 +01:00
Javier Goizueta
c8a1ef6f68
Slight optimization of gridded clustering
2015-12-22 15:25:21 +01:00
Rafa de la Torre
564ab75d2d
Use _cdb_estimated_extent instead of ST_Extent
...
With a 3.8M points table, this was a ~30% gain in my local env.
2015-12-21 18:41:50 +01:00
Rafa de la Torre
5010109c7d
Add _cdb_estimated_extent to get the extent from stats
2015-12-21 18:41:45 +01:00
Rafa de la Torre
e60f73a31b
A bit of internal documentation
2015-12-21 13:16:57 +01:00
Javier Goizueta
415a09392f
Gridded clustering aggregation strategy for overviews
2015-12-16 17:45:36 +01:00
Javier Goizueta
e5cc9ef0bd
Fix: typo
2015-12-16 17:26:17 +01:00
Javier Goizueta
f7857945c1
Change feature density computation algorithm
...
Now parameterize by the number of levels to traverse and
start with the level that covers the extent of the table
with at least N*N tiles (N=4).
2015-12-16 16:39:38 +01:00
Javier Goizueta
554464e43e
Use tile-recursive computation of feature density
...
This is a more adaptive way of estimating the feature density to
determine the base Z level.
Applying technique from http://javisantana.com/2014/10/22/traversing-quadtree.html
2015-12-16 12:14:37 +01:00
Javier Goizueta
d7c8f3d7e8
Fix: overlay generation was not using the proper scale
2015-12-15 19:13:39 +01:00
Javier Goizueta
4c85d7f3ad
Compute the reference Z level for a table
2015-12-15 19:12:48 +01:00
Javier Goizueta
d0e66910a0
Sketch for new Overview-creation functionality
2015-12-15 17:36:27 +01:00
