Vector Database Management: Difference between revisions

From GRASS-Wiki
Jump to navigation Jump to search
m (added link to progman)
 
(3 intermediate revisions by 2 users not shown)
Line 59: Line 59:
===Database Support===
===Database Support===


==== AsciiText (.csv, etc.) ====
Since GRASS GIS 7.x, SQLite is the default (local) DB driver used for GRASS vector attribute management.  


The GRASS 6 vector engine supports the following databases:
==== SQLite ====
* {{cmd|v.in.ascii}} module help page
* {{cmd|v.out.ascii}} module help page


If your .csv file contains "quoted" strings containing commas, you can use the [[AddOns#Miscellaneous_Add-ons|csv_dequote.pl]] script to parse them into a less-problematic form ready for import with {{cmd|v.in.ascii}}.
SQLite is a local database format, but much more featureful than DBF (see below). It basically combines the power of real SQL databases with the advantage of local data storage (no server needed). A nice tool to directly work in the SQLite database is [https://sqlitebrowser.org/ SQLite Database Browser].
Since GRASS 7, SQLite is the default (local) DB used for the vector attribute management.


==== DBF ====
* GRASS {{cmd|grass-sqlite}} SQLite driver help page
In GRASS 6, DBF is the default (local) DB used for GRASS vector attributes. It is easy to use but with the simplicity comes limited features. Such limits are 10 chars per column name and no support for SQL calculations in SELECT statements.
* [http://www.sqlite.org SQLite Homepage]


* GRASS {{cmd|grass-dbf}} DBF driver help page
==== ASCII text (.csv, etc.) ====


==== SQLite ====
ASCII vector data exchange:
* {{cmd|v.in.ascii}} module help page
* {{cmd|v.out.ascii}} module help page


SQLite is another local database format, but much more featureful than DBF. It basically combines the power of real SQL databases with the advantage of local data storage (no server needed). A nice tool to directly work in the SQLite database is [http://sqlitebrowser.sourceforge.net/ SQLite Database Browser].
If your .csv file contains "quoted" strings containing commas, you can use the [[AddOns#Miscellaneous_Add-ons|csv_dequote.pl]] script to parse them into a less-problematic form ready for import with {{cmd|v.in.ascii}}.
In GRASS 7, SQLite is the default (local) DB used for GRASS vector attributes.
 
* GRASS {{cmd|grass-sqlite}} SQLite driver help page
* [http://www.sqlite.org SQLite Homepage]


==== MySQL ====  
==== MySQL ====  
Line 88: Line 85:
* GRASS {{cmd|grass-pg}} PostgreSQL driver page
* GRASS {{cmd|grass-pg}} PostgreSQL driver page
* [http://www.postgresql.org PostgreSQL Homepage]
* [http://www.postgresql.org PostgreSQL Homepage]
==== DBF ====
In GRASS 6, DBF was the default (local) DB used for GRASS vector attributes. It is easy to use but with the simplicity comes limited features. Such limits are 10 chars per column name and no support for SQL calculations in SELECT statements.
* GRASS {{cmd|grass-dbf}} DBF driver help page


==== FileMaker Pro ====
==== FileMaker Pro ====
Line 166: Line 168:
* {{cmd|vector}} module help pages
* {{cmd|vector}} module help pages
* GRASS {{cmd|sql}} query help page
* GRASS {{cmd|sql}} query help page
* [https://grass.osgeo.org/programming8/vectorlib.html Vectorlib in the GRASS GIS 8 Programmer's Manual]
* [http://grass.osgeo.org/grass57/tutorial/links.html SQL reference links]
* [http://grass.osgeo.org/grass57/tutorial/links.html SQL reference links]
* [http://dev.mysql.com/doc/mysqld-version-reference/en/mysqld-version-reference-reservedwords-5-0.html MySQL reserved words list]
* [http://dev.mysql.com/doc/mysqld-version-reference/en/mysqld-version-reference-reservedwords-5-0.html MySQL reserved words list]
Line 172: Line 175:
[[Category:Documentation]]
[[Category:Documentation]]
[[Category:Vector]]
[[Category:Vector]]
[[Category:database]]

Latest revision as of 07:37, 10 October 2023

This page is a work in progress.
Please contribute if you have experience with anything that is still poorly documented.

Looking for vector geometry management? See here

GRASS GIS vector management model

The GRASS GIS vector management model allows to link map objects to database management systems (DBMS). In this short section, we try to give an overview on how GRASS matches map features and attributes.

Consider creating a map where you would handle (i) parcels with various crops, (ii) a set of owners for these parcels, and (iii) paths to access them. The latter are determined by limits between contiguous fields.

One could decide to store data in several distinct maps, e.g.:

  • a map "path" containing lines, to describe paths;
  • a map "crop" containing areas, to describe crops;
  • a map "owner" containing areas, to describe owners.

We can point several drawbacks to this method:

  • we induce redundancy of geometric features between maps "crop" and "owner". Even though their attributes are independent, areas share the same geometry;
  • topological relation between paths and parcels limits won't be ensured when it's about to modify the shape of parcels.

The concept of layers

A better solution will resort to the concept of layers. A single vector map will store geometric features, while several tables in the DBMS will handle attribute contents. Figure 1 shows how it works. Let's comment this sketch.

cats and layers in GRASS GIS
Figure 1

This vector map contains boundaries and centroids, that determine areas (see vectorintro). Boundaries are used as linear features to hold tracks geometry. Each feature is automatically assigned an internal identifier as soon as it is added to the map. The user cannot edit this id, GRASS GIS handles it for you, and ensures the uniqueness of each id. Here, ids 1 to 12 are assigned to boundaries, ids 13 to 17 correspond to centroids.

We could consider using this key-id to directly point at remote attribute records in a table. But GRASS provides a more flexible method, based on the ability :

  • to give either a single identifier to each feature or to group similar features by giving them all a common identifier;
  • to give several identifiers to an object if it is related to multiple thematic data.

These identifiers are named categories, they are organised in as many sets as you wish, that are named layers.

The concept of categories

In this example (and above sketch) objects are categorized according to three layers. Boundaries have categories in layer 1, centroids have categories in layers 2 and 3. Note :

  • 3 boundaries share category 5 within layer 1;
  • among others, object with id 7 has zero category.

Command v.category allows to maintain vector categories.

Layer 1 connects the map with a table named "paths", "cat" is the key column for this relation. The name of the key column may be different, e.g. "code" for layer 2 and table "plots". Note :

  • layer 3 does not refer to any table in the database;
  • "code" value 34 within table "plots" has no corresponding category in layer 2;
  • category 14 in layer 1 does not refer to any "cat" value in table "paths";
  • database contains a table named "owner" which is not related to any layer.

Database management in GRASS GIS is provided by the db.* set of commands :

v.to.db is the command that allows to populate a table attached to a given vector layer; not only categories can be uploaded, but geometric data too (area, perimeter, ccordinates, etc.).

Vector attribute data processing

Database Support

Since GRASS GIS 7.x, SQLite is the default (local) DB driver used for GRASS vector attribute management.

SQLite

SQLite is a local database format, but much more featureful than DBF (see below). It basically combines the power of real SQL databases with the advantage of local data storage (no server needed). A nice tool to directly work in the SQLite database is SQLite Database Browser. Since GRASS 7, SQLite is the default (local) DB used for the vector attribute management.

ASCII text (.csv, etc.)

ASCII vector data exchange:

If your .csv file contains "quoted" strings containing commas, you can use the csv_dequote.pl script to parse them into a less-problematic form ready for import with v.in.ascii.

MySQL

PostgreSQL

DBF

In GRASS 6, DBF was the default (local) DB used for GRASS vector attributes. It is easy to use but with the simplicity comes limited features. Such limits are 10 chars per column name and no support for SQL calculations in SELECT statements.

FileMaker Pro

William Kyngesburye wrote on the grass-user mailing list:

I think the key to access FileMaker DBs from GRASS is that the FileMaker ODBC connector is for OSX's iODBC, not UnixODBC.

GRASS 6 has an iODBC configure option. It's the same --with- odbc-* options, it just tries iodbc if it can't find unixodbc.

Oracle

This has been reported to work. Try ODBC and search the mailing list archive. The connection is done via OGR.

ODBC

  • External DB support via ODBC (e.g. FileMaker Pro)
  • unixODBC is required to make it work
  • to configure you could use the graphical frontend ODBCConfig to configure your ODBC connection.

Example-entry in ~/.odbc.ini for usage of ODBC with PostgreSQL

  [dbname]
  Description         = PostgreSQL database for my project
  Driver              = postgres
  Trace               = No
  TraceFile           =
  Database            = mydb
  Servername          = myserver
  UserName            = myusername
  Password            = mysecretpasswd
  Port                = 5432
  Protocol            = 9.0.3
  ReadOnly            = No
  RowVersioning       = No
  ShowSystemTables    = No
  ShowOidColumn       = No
  FakeOidIndex        = No
  ConnSettings        =

Additionall you need to define the libraries to use for the different drivers in /etc/odbcinst.ini.

  [postgres]
  Description     = ODBC for postgres
  Driver          = /usr/lib/unixODBC/libodbcpsql.so
  Setup           = /usr/lib/unixODBC/libodbcpsqlS.so
  FileUsage       = 1

Concepts and jargon

  • GRASS 6 Terminology
  • Table
    • Table column
    • Table row
    • Vector map layer
    • Each vector file has a special data field named "cat" (derived originally from "category"), filled with integers, that serves to identify each vector object. The 'cat' field also serves as a "key field" that can link each vector object with a corresponding record in an attributes table of a database (NB: 'cat' values do NOT have to be unique for vector objects, but DO have to be unique in an attributes table, permitting both one-to-one and many-to-one relationships). The attributes table must contain a key field, filled with integers (only integers are permitted to serve as key fields in GRASS), that matches the values in the vector 'cat' field.
    • A vector can be linked with more than one attribute table, using LAYERS. By default, every vector file has a LAYER 1 with a 'cat' field, filled with integers to identify each vector object. Additional *LAYERS*, along with their associated 'cat' fields, can be created using v.category. Each LAYER has its own independent 'cat' field that can be used to link with a separate attributes table.
    • For example, a vector file of cities can have *LAYER 1* whose 'cat' field links the vector points with an attributes table of demographic data; it can also have a LAYER 2, with an associated 'cat' field (independent of the 'cat' field of LAYER 1), linked to a different attributes table of economic data. Values in the 'cat' fields of different LAYERS can be the same or different. The 'cat' values and linked attributes table can be queried independently for each LAYER. In the example, one can query the cities by population from the demographics attributes table linked with the 'cat' field of LAYER 1, or query the cities by household income from the economic attributes table linked with the same vector points through the 'cat' field of LAYER 2. _MichaelBarton - 12 Nov 2005_

Common tasks

Background info find in vectorintro Vector data processing in GRASS GIS.

More Help