Fix plotting with custom types as keys

[nvaytet]

This used to raise with a strange error:

import scipp as sc
import plopp as pp

a = pp.data.data1d()
b = a * 2

from typing import NewType
A = NewType("A", str)
B = NewType("B", int)

pp.plot({A: a, B: b})

This is a common use case when computing multiple results from a sciline workflow, as it returns a dict with custom types as keys.

[SimonHeybrock]

Is this actually the correct place for the string conversion, or would it be better to accept, e.g., any Hashable as name and add string-formatting where actually needed?

[nvaytet]

I did some more digging and I think this is the right place for the conversion.
It is not about string formatting, but how we handle different input types to the Node constructor.

The error that was being raised before the fix is

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[3], line 8
      5 A = NewType("A", str)
      6 B = NewType("B", int)
----> 8 pp.plot({A: a, B: b})

File [~/code/path/plopp/plotting/plot.py:173](http://localhost:8965/home/nvaytet/code/path/plopp/plotting/plot.py#line=172), in plot(obj, aspect, autoscale, cbar, clabel, cmap, cmax, cmin, coords, errorbars, figsize, grid, ignore_size, legend, logc, logx, logy, mask_cmap, mask_color, nan_color, norm, scale, title, vmax, vmin, xlabel, xmax, xmin, ylabel, ymax, ymin, **kwargs)
    171 ndims = set()
    172 for n in nodes:
--> 173     ndims.add(n().ndim)
    174 if len(ndims) > 1:
    175     raise ValueError(
    176         'All items given to the plot function must have the same '
    177         f'number of dimensions. Found dimensions {ndims}.'
    178     )

File [~/code/path/plopp/core/node_class.py:89](http://localhost:8965/home/nvaytet/code/path/plopp/core/node_class.py#line=88), in Node.__call__(self)
     88 def __call__(self) -> Any:
---> 89     return self.request_data()

File [~/code/path/plopp/core/node_class.py:147](http://localhost:8965/home/nvaytet/code/path/plopp/core/node_class.py#line=146), in Node.request_data(self)
    144 if self._data is None:
    145     args = (parent.request_data() for parent in self.parents)
    146     kwargs = {
--> 147         key: parent.request_data() for key, parent in self.kwparents.items()
    148     }
    149     self._data = self.func(*args, **kwargs)
    150 return self._data

File [~/code/path/plopp/core/node_class.py:149](http://localhost:8965/home/nvaytet/code/path/plopp/core/node_class.py#line=148), in Node.request_data(self)
    145     args = (parent.request_data() for parent in self.parents)
    146     kwargs = {
    147         key: parent.request_data() for key, parent in self.kwparents.items()
    148     }
--> 149     self._data = self.func(*args, **kwargs)
    150 return self._data

TypeError: _typing._idfunc() takes exactly one argument (0 given)

This happens because when we take in a dict of objects, we send each object through a pre-processor node which converts the input to a data array. One of the args of that pre-processor is the name of the data array (which is used for legends or axes labels).

The name gets wrapped into a Node itself because it is an input to the pre-processor node. When wrapping the name, the Node constructor checks if the thing it is wrapping is callable. The types A and B are callable and so they get handled as if they were a function that returns data, when in fact they should either be converted to a string, or just wrapped in a lambda function.

The only reason why it worked before with pp.plot({int: a, float: b}) and not with pp.plot({A: a, B: b}) is that int() yields 0 and float() yields 0.0 (i.e. it is possible to call int and float with no arguments, while calling A() and B() fails).

[jokasimr]

Is the example in the PR description correct?

I don't understand what the figure should display.

# The values A and B are types, plopp can not plot types.
pp.plot({str: A, int: B})

# Did you mean:
pp.plot({A: a, B: b})

[nvaytet]

Did you mean:
pp.plot({A: a, B: b})

Yes! (updated)