forked from postgis/postgis
-
Notifications
You must be signed in to change notification settings - Fork 0
/
MIGRATION
182 lines (140 loc) · 7.18 KB
/
MIGRATION
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
PostGIS 2.0 Migration Guide
===========================
PostGIS 2.0 introduces changes that may affect backwards compatibility
for some applications. This document lists those changes.
Deprecated Functions Removed
----------------------------
Since PostGIS 1.2, we have been encouraging users to use functions with
the ST_ prefix. The ST_ prefixed functions are part of the ISO SQL/MM
standard and used by other spatial SQL databases, so we want to conform
as closely to the industry usage as we can.
At 2.0, we have finally removed the old functions. If you have an
application that *still* requires the old function signatures, please
look to the legacy_*.sql files to add the old signatures back in.
If you are using MapServer, upgrade to the latest versions of the 6.0
or 6.2 releases. For earlier releases, you will have to install the
legacy.sql files.
Unknown SRID
------------
The "unknown SRID", assigned to geometries when SRID is not specified,
is now 0, not -1. ST_SRID will now return 0 when called on a geometry
with no known SRID.
ST_AsBinary, ST_AsText and 3D
-----------------------------
Previous versions of PostGIS returned 2D versions of objects when
ST_AsBinary and ST_AsText were called. PostGIS 2.0 returns ISO SQL/MM
versions when 3D or 4D objects are used. For example, a 3D point with a
Z dimension will return as 'POINT Z (0 0 0)'. A 4D point will return as
'POINT ZM (0 0 0 0)'.
For WKB, extra dimensionality is indicated in the type numbers. The Z,
M, and ZM type numbers are indicated by adding 1000, 2000, and 3000
respectively to the usual 2D type numbers. So, for example, a 2D Point
has a type number of 1. A 4D PointZM has a type number of 3001. A 2D
polygon has a type number of 3. A PolygonZ has a type number of 2003.
geometry_columns
----------------
In PostGIS 2.0, the ``geometry_columns`` metadata has been changed from
a table to a view.
* For applications that manage ``geometry_columns``
using the standard AddGeometryColumn() and other methods, there should
be no change.
* Applications that have inserted directly into ``geometry_columns``
will need to either move to the standard functions, or change
their create table syntax to include type/srid/dimensionality
information.
For example, to specify a 3D feature in WGS84::
CREATE TABLE mytable (
id SERIAL PRIMARY KEY,
g Geometry(PointZ,4326)
);
* Applications reading metadata from ``geometry_columns`` should expect
that the geometry type strings will be mixed case and include the
dimensionality modifiers. For example: PointZM, MultiLineStringZ.
* If you have views built on tables that were built
the old constraint way or that use processing
functions to return new geometries, they will not register correctly
in geometry_columns. They will all appear as Geometry with srid=0.
In order to fix this, you have to cast the geometries in the view.
So for example, If you had a view that looked like this:
CREATE OR REPLACE VIEW vwparcels AS
SELECT gid, parcel_id,
ST_Transform(geom, 26986) As geom_26986, -- a transformed geometry
geom, --one that used constraints before
ST_Centroid(geom) As cent_geom -- one that used processing function
FROM parcels;
You will need to drop and recreate the view.
Sadly CREATE OR REPLACE will not work since the casting
makes geometries be treated as a different type as far as
CREATE OR REPLACE is concerned.
NOTE for this, it might help to lookup in geometry_columns
to see the raw table srid/type
Alternatively you can convert some of your table columns to typmod,
then you would only need to CAST when doing processing.
This however still requires you rebuild your dependent views
DROP VIEW vwparcels;
CREATE OR REPLACE VIEW vwparcels AS
SELECT gid, parcel_id,
ST_Transform(geom, 26986)::geometry(MultiPolygon,26986)
As geom_26986,
geom::geometry(MultiPolygon, 2249) As geom,
ST_Centroid(geom)::geometry(Point,2249) As cent_geom
FROM parcels;
Restoring data from prior versions -- GOTCHAS
-----------------------------------------------
The populate_geometry_columns() function in PostGIS 1.5 and below,
used the deprecated functions ndims() and srid() for defining constraints.
As a result these tables will not
restore into a fresh PostGIS 2.0 database unless you
1) Drop these contraints before you restore the data
or
2) Install the legacy functions srid() and ndims() found in the legacy.sql
file before you try to restore these tables.
Function is not unique after restore
-------------------------------------
After you restore old data into a fresh PostGIS 2.0 database,
YOU MUST install the uninstall_legacy.sql
file after the restore. If you don't you will run into
"function is not unique" errors in your applications.
The other issue with having the old functions is the old functions you
restore will be bound to the old postgis library which is incompatible
with the new postgis geometry structures.
If you still require legacy functions, you can install the legacy.sql
file after the uninstall_legacy.sql.
GOTCHA with uninstall_legacy.sql
--------------------------------
The uninstall_legacy.sql will not be able to uninstall functions and will
output errors detailing objects using these functions that are currently
being used in views and sql functions.
If you run it in PgAdmin, the whole script will rollback not
uninstalling anything. So you should run it in PSQL mode.
If you get errors in uninstall_legacy.sql -- convert those views and sql
functions to use the newer equivalent functions
and then rerun the script again.
BENCHMARK NOTES
------------------------------
Querying a whole geometry_columns table of 200-240 someodd records
(tables/views all constraint based)( On windows 9.1beta2)
Speed for this set of data I'm okay with though a bit slower
I think the main issue why fancy is slower is that to get the fancy name
I need to do a double call to get dims since fancy_name needs dims as well
This I expect to be much faster using typ_mod
SELECT * FROM geometry_columns_v; -- 2063 ms, 1063 ms (using fancy name)
SELECT * FROM geometry_columns_v2; -- 656 ms (this is not using fancy name)
SELECT * FROM geometry_columns; -- original static table 31ms - 150 ms
-- Doing single selects of a table
-- constraint based one -- 76ms, 46ms
SELECT * FROM geometry_columns_v where f_table_name = 'streets';
-- One I converted to typmod -- 76 ms, 48 ms (first call was slower)
SELECT * FROM geometry_columns_v where f_table_name = 'buildings_1995';
--Filter query for schema --
-- these are constraint based (7 rows) -- 76 ms, 240 ms, 68 ms
SELECT * FROM geometry_columns_v where f_table_schema = 'tiger';
-- CONVERTING TO Typmod as I feared is slow for tables with data.
I think it follows the same model
-- as converting varchar(100) to text or varchar(100) to varchar(300) etc.
-- This maybe can be remedied somehow in 9.1 since it hasn't been released
-- (haven't tried doing this exercise in 8.4, 9.0)
-- This took 36,018 ms for table of 350,572 records.
I assume it will be linear with number of records.
ALTER TABLE buildings_1995
ALTER COLUMN the_geom TYPE geometry(MultiPolygon,2249);