GSoC 2016 PyQt based GUI of GRASS automatically generated from XML

Title: PyQt based GUI of GRASS automatically generated from XML
Student Name: Ondrej Pesek, Czech Technical University in Prague
Organization: OSGeo - Open Source Geospatial Foundation
Mentor Name: Vaclav Petras, Anna Petrasova
GSoC proposal: view proposal
Repository: github


GUI forms (module dialogs) are now automatically generated from XML into wxPython. My idea is to create code that is generating the same code into more powerful Qt.


Create a table of individual widgets, their features and when their are used. Existing GUI in GRASS GIS and also in QGIS GRASS plugin and QGIS Processing plugin are good starting points.

List of standard options in GRASS GIS gives list of commonly used parameters which are likely to become specialized widgets (but it is not 1:1 - a lot of them are actually the same map selection widget just with different parameters). The general specification of options and flags is in the programming manual

There is more ways to do it. We will try to make more files. Nowadays, it is just one file with more than 3000 lines. It is not very readable. We hope that we can change it.


In the first version, creating GUI will work just as script started from GRASS. Of course I want to implement it as normal function (there is variable "gui" in GRASS), but it's not necessary. When the GUI is ready and compatible with many functions, then tere is possibility to make it. (not in timeline, additional goal)

The same is with styling. I will try to make everything very similar to current GUI, but I didn't make any window in timeline for that. I will work on that within all the GSoC period assuming that I have the time.

I have a problem with the end of the june. I have Final Examinations at my school so maybe I won't have so much time. I will try to make everything here ahead of time. In the case I won't, I will definitely catch up it after this period. It will be same for every hiatus that can raise (two last weeks of GSoC).


timeline created during bonding
Community bounding period Familiarization with the existing code *
Let the summer begin
MAY 23 - 27 Creating empty generated GUI *
MAY 30 - JUNE 3 Creating basic GUI (just basic widgets) of some function - e.g. r.buffer *
JUNE 6 - 10 Updating GUI - using standard Qt widgets and creating the first version of copyable/runable code *
JUNE 13 - 24 Updating GUI - using standard Qt widgets + make Run button runable *
Mid term evaluation
JUNE 27 - JULY 8 Implementation of custom widgets (gselect.Select), running also flags *
JULY 11 - 22 Linking widgets, dependencies (select map -> columns) *
JULY 25 - 29 Clearing of code *
AUGUST 1 - 5 Testing *
AUGUST 8 - 13 Writing documentation *
The end




Weekly reports

May 23 - 27

Designed basic GUI shuck. From XML is now automatically generated GUI with name, keywords and basic layouts (description, tabs, buttons).

Code now works as a script with parameter - for example r.buffer

May 30 - June 3

I have made GUI using some basic widgets - checkboxs, labels, lineedits...

Now the code is reading those types - float, range, sql_query, flags

It also writes the label (or name) and type into GUI.


June 6 - June 10

I have upgraded GUI - now it works also with the string and integer types. (on the basic level)

I also have made some changes - SqlQuery is now lineedit and when the float isn't multiple, the GUI generates QDoubleSpinBox.

The important thing is the first version of copyable/runable code. The buttons don't work yet, but user can see it and it automatically reads the code.

And one elegant thing - all the widgets are now stretched upper.

June 13 - June 17

The Run button really runs (but still not with flags). It's still the only button with any function. Also the code-string on the bottom of GUI was recreated. Now it's just one (only readable) textedit. When the string is too long, you will get the scrollbar.

Completely retyped code in Now it works like Factory. Factory is dynamically creating the list of methods classes with canHandle method and the called class depends on the values in this method. It means that there are not so many "ifs" and the code is better readable. And those classes return "self" instead of method get().

Float coords is lineedit.

When the string has values, it's editable combobox with implemented values.

Nearly last thing: I implemented the TreeComboBox for inouts/outputs. Both raster and vector. It works very similarly to the one in old GUI. Mapset is parent and then you have many children to choose. It's in and in is just the class inheriting it from gselect with added canHandle method. It's much readable for anyone than just directing program elsewhere.

For not implemented thing I use lineedit (because You can write there anything manually). Now I highlight it with red colour and word TODO in GUI. It's just for me and will not be in the real version.

I'm sorry but at June the 27th, I have my bachelor final exams at school so I won't be able to do anything related to GSoC in next week (I really have to learn something before exams). I have done everything in my timeline for next week so I hope it's okay. Thanks.

June 20 - June 24

Due to my final exams, as I said, I didn't do anything this week.

June 27 - July 1

The Run button works now also with flags. Also the user can see flags in string below.

For prompt='file' I implemented lineedit with Browse button to choose file.

July 4 - July 8

New widgets: cats, multiple values, raster_3d

Little changes in other widgets: Mapset is not selectable in treewidget, flags and default widget were moved into its own classes

If widgets have both label and description, description is used as help/tooltip

Massive refactoring of code for command (running, changing dictionary, changing cmd string that user see)

July 11 - July 15

New widgets: layer, columns, multiple values with values_desc, separator, group, colors

helpButton and closeButton work

Help/tooltip also for flags

Fix: Float or int equals 0 -> delete from cmd, more flags than one will not call overwrite, set minimum size to browsebutton

Refactoring (using methods as callbacks)

July 18 - July 22

New widget: dbtable

I had finally ordered tabs. Required tab is the first one and Optional tab the last one. And if widget has defined guisection, it is now prior to Required.

I had fixed problems with spinboxes. They now allow you to input number greater than 100 and also negative numbers.

Was created first version of sizepolicy. Dialog width and height still depends on content volume, but now just till some limit. After that, the tab gets sliders to scroll.

July 25 - 29

New widgets: Mapsets, locations and dbtable (created dependencies between those), special widget for quiet/normal/verbose module output

Created definitive (I hope) size policy. ialog width and height still depends on content volume, but now just till some limit. After that, the tab gets sliders to scroll.

Fixes and updates: Widgets Columns and Layers now work with map in widget 'map' when 'input' does not exist; fix widget Colors for default value in 'r:g:b' style

Main window and dialogs now have grass logo and icon

Required widget is highlighted by red star

Tab 'required' isn't visible when it's empty

Was created documentation for .py files and for individual class, methods...

key_desc used instead of type (when showing widget name)

Big refactoring using default pep8

I'm sorry but I will be now for two weeks (as I said at the beginning) off. Off the internet, off the computer. I should be able to work one day in next week, I hope. I hope that I will finish icons widget...

July 1 - 5

New widget for choosing icon/symbol from special dialog (see in d.vect)

I'm sorry but I will be now for whole week off. Off the internet, off the computer. I will not be able to do anything in GSoC.

July 8 - 12

Wiki page:

A link to my commits:

The goal of the project was to create new GUI for modules based on PyQt. The old one is based on wxPython. The reason is that there are some bugs in wxPython and PyQt is now much more powerfull. The second reason is that the code of old GUI is written bad way. It's almost impossible to read what something do and work with that. So the second goal was try to write higher-quality code.

My project is now usable. It's interface is mostly based on the old one to avoid the need for studying new approach (the code isn't), but some widgets were changed due to my and my mentor's opinion. For example we use spinbox for int widgets or there is completely new widget for choosing quiet/normal/verbose module output.

What have been done is visible in github repository. Screenshots of widgets and other stuff is at wikipage.

Last modified 14 months ago Last modified on Aug 14, 2016 3:30:25 AM

Attachments (19)

Download all attachments as: .zip