I really like the histogram creation in the statistics node. My current use case involves investigating the distribution of sampled values from various distributions covering the range 0-1.
It would be really useful if I could specify the x axis min and max values so that the scale for all the histograms generated is the same.
Probably it would be better if the domains could be used as the min/max values in these cases -if those values are present and a configuration option is selected. The domains can individually adjusted and would cause problems only if something is outside of the set min/max values, but the domain calculator would give reasonable defaults. (In my opinion the current behaviour might remain the default and a checkbox should be checked to use the domain values.)
In your version, what should happen with the values outside the predefined range? (Should they be handled similarly to missings, NaNs and infinite values, or count them to the infinite values? Or fail the execution?)
Indeed there are factors to consider outside of my specific example (and more in depth in my example).
The domain of the columns would be a good choice for min and max values as a general case but wouldn't necessarily solve my problem. I may have only returned sampled values below 0.8 for a specific column, I would still want to plot the X axis up to 1 as this is the range I could have recived values from. The domain wouldn't reflect this?
In the example I gave if values fell outside the 0-1 range something has gone horribly wrong and I would want it to notify me that the range doesn't cover the values. For example it could check that the user defined values are not larger than the domain min and smaller than the domain max. This would prevent the choise of x axis limits but a user may may only be wish to display part of a histogram and a warning may be more appropriate?
I may have only returned sampled values below 0.8 for a specific column, I would still want to plot the X axis up to 1 as this is the range I could have recived values from. The domain wouldn't reflect this?
The Edit Numeric Domain node could handle this case (and probably adding two rows with 0 and 1 there, calculating the domain and filtering out the new rows would act similarly), though if you do not need extra control your proposed solution would be easier to understand.
(Another option would be using predefined properties to adjust the axis ranges. [The SVG preferred size already uses this for visualization.] If they are not present, the computed min-max are used, if present, the set values. Though one can argue that this might be redundant as the domain might have similar purpose.)