Pipeline representation improvement ideas

[YooSunYoung]

I came across with some problems and requirements while I was trying to use sl.Pipeline html wrapper in scipp/essnmx
documentation.

Wrong name parsing of NewType in python 3.9 I just saw that this issue was marked as wontfix...! Sorry

• Dict-like representation of Pipeline. So that it's easier to turn them into a html wrapper or other formats.
• Pipeline.__repr__. It should be easy if the Dict-like representation is implemented.
• Customize Value representation. For example, when I have a function it shows the address of the instance not the name of the module and the function. It'll be useful to pass a customizing function by the type.
  i.e.

import sciline as sl
pl = sl.Pipeline()
pl[int] = 0
pl.repr_value_renderer = lambda x: f"number: {x}"
print(pl)
# [ 
#  { 'Name': 'int', 'Value': 'number: 0', 'Source': '' }
# ] 

I'm suggesting this because, we can't handle all different types of providers and values on our own. However, if the Pipeline wrapper doesn't show what I want to show, it's less useful in the documentation.

Maybe it should be discussed together with scipp/scipp#3305

[SimonHeybrock]

Item one is a wontfix, for the others, does Pipeline._repr_html_ not do the job for you?

[YooSunYoung]

It mostly do, but these ideas started only because nmx workflow has function object as a parameter.
I didn't want to add instance address in the documentation...
And we have __repr__ for most of important data structures, so... why not for Pipeline...?
It'll make logging much easier.

[jl-wynen]

Item 4 should be improved by the new utils in #124