Refactoring unit implementation for use in dataframes and plots

[DunklesArchipel]

The data frames in the database do not contain any information on the currently used units. We create the data frame from the CSV file, which has been created with SI units.

In our previous version, the data frame could be created with units different from SI. However, the returned df obviously does not contain this information.

Such information is crucial for the plot module, where the units should be printed on the x and y-axis labels.

Another issue is that the y axis can be either a current (I) or a current density (j) with distinctly different units.

x and y units are now returned by xy_units(). It is possible to return SI units, custom units or the original units from the figure.

Maybe there is a smarter quick and dirty solution than this one. Otherwise, we could wait for pandas 2.0 where unit storage along with data frames is eventually implemented.

Also the rendering of the Figure on the cv_entry.md has been updated. (Not yet super nice, but better than before.)

[saraedum]

Do you have a link for that pandas update? I also would not mind to use an experimental change in pandas here.

[saraedum]

Do you have a link for that pandas update? I also would not mind to use an experimental change in pandas here.

I could not find any actual pandas 2 implementation. It seems to be just a rough roadmap.

[saraedum]

Have you had a look at astropy time series? They support quantities, see https://docs.astropy.org/en/stable/timeseries/pandas.html.

[saraedum]

I proposed a way to get rid of the I/j hack. Otherwise, this looks great apart from some wording issues, e.g., I am pretty sure that the lines in docstrings that end in :: should not end in a ..

[saraedum]

Why don't split this into x_unit and y_unit? There seems to be no interaction between the two parts of this method.

[saraedum]

This if-j-if-I is really annoying.

Could we do the following: We have x_unit and y_unit which return the units of the axes, as you planned here. Then we have x and y which return the name of the axes, i.e., U and I/j.

Suggested change

	if 'j' in column_names:
	yunit = u.A / u.m**2
	if 'I' in column_names:
	yunit = u.A
	if self.y == 'j':
	yunit = u.A / u.m**2
	elif self.y == 'I':
	yunit = u.A
	else:
	raise NotImplementedError("Unexpected naming of y axis.")

Also, we should always catch unexpected values as early as possible, as this makes debugging much easier when something goes wrong.

[saraedum]

Suggested change

	if 'j' in df.columns:
	fig.add_trace(plotly.graph_objects.Scatter(x=df['U'], y=df['j'], mode='lines'))
	ylabel = f'j [{str(yunit)}]'
	if 'I' in df.columns:
	fig.add_trace(plotly.graph_objects.Scatter(x=df['U'], y=df['I'], mode='lines'))
	ylabel = f'I [{str(yunit)}]'
	fig.update_layout(template="simple_white", showlegend=False, autosize=True, width=600, height=400,
	margin=dict(l=70, r=70, b=70, t=70, pad=7),
	xaxis_title=f"U [{xunit}]",
	yaxis_title=ylabel)
	fig.add_trace(plotly.graph_objects.Scatter(x=df[self.x], y=df[self.y], mode='lines'))
	fig.update_layout(template="simple_white", showlegend=False, autosize=True, width=600, height=400,
	margin=dict(l=70, r=70, b=70, t=70, pad=7),
	xaxis_title=f"{self.x} [{xunit}]",
	yaxis_title=f"{self.y} [{yunit}]")

[DunklesArchipel]

@saraedum I hopefully implemented all your changes.

The doctests fail probably due to the latest update on svgdigitizer that you did. I cannot reproduce that particular doctest error locally, even with the version of the svgdigitizer.

[saraedum]

@saraedum I hopefully implemented all your changes.

The doctests fail probably due to the latest update on svgdigitizer that you did. I cannot reproduce that particular doctest error locally, even with the version of the svgdigitizer.

Yes, I think that's the problem described at https://github.com/echemdb/website/blob/main/environment.yml#L30.

To reproduce, I think you have to delete the generated/ directory somewhere that holds a local cache of digitized data.

Should be fixed by #96.

[saraedum]

There are merge conflicts @DunklesArchipel .

[linuxrider]

Looks good to me.

[saraedum]

Looks great. I have just some minor comments. Feel free to merge once addressed.