v.dissolve: Compute attribute aggregate statistics

[wenzeslaus]

In addition to geometry dissolving, compute aggregate statistics for the attribute values of dissolved features with v.db.univar and SQL.

v.db.select with group is used to obtain unique values of the column the dissolving is based on. Add column and update now happens for every value, column, and statistics.

Originally implemented with v.db.univar only because it has a good set of functions, but direct SQL is faster and potentially can have more functions (although default SQLite has less).

Auto-generates names and combinations of column-method for convenience, but when all needed parameters are provided, uses them as is.

Has documentation, examples, image for original functionality, and test (image generated in notebook).

Uses plural for columns and methods.

Removes duplicate columns and methods for non-explicit automatic (interactive) result column handling.

Support SQL expressions as columns (as in v.db.update query_column or v.db.select columns). Supports general SQL syntax just like v.db.select for the price of less checks. Supports also text-returning aggregate functions and functions with multiple parameters such as SQLite group_concat. Supports any layer, not just 1, for attributes.

Uses a simple SQL escape function to double single quotes.

Requires v.db.univar JSON output and v.db.select column info in JSON output.

Handles cleanup from the main function. Removes global variables. Uses PID and node name for the temporary vector. Partially modernizes the existing code by using gs alias instead of grass alias. Improves author lists.

[wenzeslaus]

This depends on JSON output in #2386.

[wenzeslaus]

Notebook to test: https://mybinder.org/v2/gh/wenzeslaus/grass/v_dissolve-attr-stats?urlpath=lab%2Ftree%2Fscripts%2Fv.dissolve%2Fv_dissolve.ipynb

[HuidaeCho]

@wenzeslaus Overall, it looks good to me. Just have a few minor comments.

[wenzeslaus]

I made some significant updates. The interface is now tailored for two different use cases, one for interactive use, when a lot of things happen automatically and then the other one for scripting when user is expected to be very explicit about what should be computed. This is now described in the documentation.

I resolved some of the comments after writing a comment with resolution, but some may still require some discussion.

Thank you for the feedback, @HuidaeCho.

[wenzeslaus]

Should the option names be in singular or plural? For this module, it is aggregate_column, aggregate_method, and result_column versus aggregate_columns, aggregate_methods, and result_columns.

Is it singular as in:

r.patch input=aaa,bbb,ccc output=xxx
v.patch input=aaa,bbb,ccc output=xxx
g.list type=raster,vector

or plural as in:

v.db.select map=aaa columns=xxx,yyy
v.db.addcolumn map=aaa columns="xxx double precision,yyy integer"
v.db.addtable map=aaa columns="xxx double precision,yyy integer"

?

Singular versus plural in option names

The following does not include more special options in terms of use of singular and plural such as cats, coordinates or GDAL options.

Standard options

Standard options with multiple set to yes.

id	name	s/p
G_OPT_V_INPUTS	input	singular
G_OPT_V_MAPS	map	singular
G_OPT_V_TYPE	type	singular
G_OPT_V3_TYPE	type	singular
G_OPT_DB_COLUMNS	columns	plural
G_OPT_R_INPUTS	input	singular
G_OPT_R_OUTPUTS	output	singular
G_OPT_R_MAPS	map	singular
G_OPT_R_ELEVS	elevation	singular
G_OPT_R3_INPUTS	input	singular
G_OPT_R3_MAPS	map	singular
G_OPT_M_DATATYPE	type	singular
G_OPT_STDS_INPUTS	inputs	plural
G_OPT_STRDS_INPUTS	inputs	plural
G_OPT_STRDS_OUTPUTS	outputs	plural
G_OPT_STVDS_INPUTS	inputs	plural
G_OPT_STR3DS_INPUTS	inputs	plural

Notes: G_OPT_V_OUTPUTS does not exist, but given the other ones, it would be output. There is many temporal options, but they are not used that much as the other ones: all combined are used 10 times while just G_OPT_V_INPUTS is used 5 times and G_OPT_R3_INPUTS 3 times.

Vector modules in C

C modules with multiple = YES in the vector directory. Other modules not listed (too many).

module	option	s/p
v.normal	tests	plural
v.clean	tool	singular
v.net.iso	costs	plural
v.what	layer	singular
v.distance	column, upload	singular
v.build	option	singular
v.in.pdal	class_filter	singular
v.in.ogr	layer	singular

Python modules

Python modules with multiple: yes.

module	option	s/p
v.db.select	columns	plural
v.db.univar, db.univar	percentile	singular
v.db.addtable	columns	plural
v.db.addcolumnumn	columns	plural
r.patch	input	singular
r.texture	method	singular
r.buffer.lowmem	distances	plural
r.semantic.label	semantic_label	singular
r.in.wms	layers, styles	plural
g.search.modules	keyword	singular
t.vect.db.select	columns	plural
t.merge	inputs	plural

[wenzeslaus]

While this could be faster or parallel, the aggregation works well for simple cases and can deal with some complex cases, too. From my perspective, this is ready to be merged right after #3090.

Issues addressed, code changed significantly, review is no longer relevant.