silx-kit / silx-training Goto Github PK

We have three binders links in the README:

Binder.org.

Not working because there is a timeout when trying to install jupyter-desktop-server from Git.

Also, it seems we should now use jupyter-remote-desktop-proxy instead.

PaNOSC fedcloud

The service seems down ? Given PaNOSC is finished, it was to be expected ?

Gesis

Same as Binder

The files contains uint, it is dangerous do computation without overflow

It is needed to present how-to read the documentation of Qt in order to write PyQt code:

• Conversion from C++ to python: ->, :: to ., etc...
• Slots are just methods in PyQt
• Qt properties are actually a getter propertyName() (isPropertyName()) and a setter setPropertyName()

Some ideas following last training:

• Make git a standalone training: to cover it better (update branch from upstream, resolve conflicts) and leave room for exercises.
• Talk more about project structure

• Normalize dark/background

People who do not know this domain had problem to see it is the same thing.
Maybe a slide about the technic can be useful anyway.

Ideas to improve the introduction to numpy:

• re-order from simple to more complex stuff
• provide realistic examples/exercises

Reordered presentation:

• basic plotting with matplotlib: plot and imshow
• create arrays + load/save .npy .txt files
• element-wise basic operations (exercise/example x - y + flat field correction (2D) or linear background substraction (1D))
• numpy functions: min, max, sum etc... + axis argument (possible exercise/example: process a stack of images to show the difference of std on detector module edges)
• numpy modules: linalg, random, fft
• indexing (+ broadcasting?) exercice: 2x2 binning
• numpy array insights: shape, strides and base
• last exercise: cos(row) * sin(col)

silx training: Add information on how to connect/start the exercice: connection to server + start jupyter or ipython

in part 1 - loop example: This would avoid loosing some time on it

Some suggestion from Loic, Wout, Olof, Alex, Armando ...

1. Summarize the many programming languages into a table, will be faster and easier for the comparison.
2. Performance portability of the code
3. Emphasize on (long term) maintainability and the relation with metaprogramming
4. Maturity
5. show the generated code to prove it is readable (?) remove the black-magic part.

Some people do not understand at all the tests.

It could be useful to introduce it with a real example.

• A real function
• The way you try to see if it is working (maybe using ipython)
• What if we change the function?
• Introduce that it could be written once and reused many
• Maybe show that it can improve the function: A failing test case, but the test is right, and the tested function have to be fixed.

https://docs.python.org/3/library/zipapp.html

• NXsubentry has to be inside NXentry and not outside as in the drawing.

Feel free to add more things:

• No instruction to copy training materials
• Typos in the images about references in lists #97
• Some cells were hidden in python notebooks 2 and 3 (too many cells in a slide)
• Numpy exercise numbers do not correspond to solutions numbers #99
• Typo in exercise 5 of numpy (100x100 1D) #99

Remove example with *args, **kwargs and simply inform them that exists.
Analysis is taking to much time and is confusing for users.