EzTemplate defaults "chop" axes when writing to file

[durack1]

@danlipsa @doutriaux1 the current EzTemplate defaults do not prevent the axes being written off the canvas, and so when an output file is written we get the below. I also note that the axes, legend etc overwrite one another:

import EzTemplate
import vcs
import numpy as np

d = np.ones([10,10])
gm = vcs.createboxfill()
canvas = vcs.init(geometry=(2700, 2700), bg=True)
M = EzTemplate.Multi(columns=1,rows=2)
# First panel
t = M.get()
canvas.plot(d,gm,t)
# Second panel
t = M.get()
canvas.plot(d,gm,t)
canvas.png('EzTemplate-broken')

It would be good to revise these templates so that by default, all axes and informational text is captured within the bounds of the output format and that overwrites do not occur

[doutriaux1]

@durack1 I don't understand the issue, all that EzTemplate does is rescaling your original template (or the default one if you do not pass one). So if your original template as axes on they will be drawn. What is the issue here?

[durack1]

@doutriaux1 see the axis labels for the y-axis? They are chopped in half and unreadable.. The labeling is also overwriting other elements in the plots - even though these are defaults.. So the colorbar labels and the x-axis for the lower plot

[doutriaux1]

I see. Ok that needs tweaking...

[doutriaux1]

@durack1 in the case of y-axis. Should we attempt to scale down the font until it fits?

[doutriaux1]

@durack1 I'm not sure what to do about variable_1 etc.. that end up on top of top plot. Should I add an option turn_off_unfit_objects if the end up being chopped or on top of others?

[durack1]

@doutriaux1 with the example above, I think the font sizes are right.. I just think the plot area needs to be made to fit inside the canvas. So I would recommend making the plot areas smaller so that the overlaps no longer occur.. The descriptive info variable_1 etc should be kept, unless a user specifies these labels are off

[danlipsa]

@doutriaux1 @durack1 If I remember correctly, I think there is a fundamental problem with vcs in that you can specify the area where the plot will be placed and then the labels are placed on the outside, regardless of the fact that they may be cut off. I think a proper fix would be to place the entire plot including the labels inside the specified area.

[durack1]

@danlipsa thanks for chiming in, yes I think there is some rethinking to be done with the templates particularly now that many of the other elements are working correctly.. I'm happy to contribute to discussions when this issue gets addressed

[jypeter]

This is just a related comment. How is a user supposed to find documentation about EzTemplate from the web site?

There is nothing about it on the site from Documentation or Documentation/Tutorials. And in Gallery/Template, there is only an example (a good one) about changing the templates by hand. Knowing about EzTemplate, I have googled "EzTemplate cdat" and found https://github.com/UV-CDAT/cdat-site/blob/master/eztemplate.md. But I don't know if this is actually somewhere on the web site

[doutriaux1]

@durack1 I get your point but EzTemplate really focuses on the data and legend area. I could add an option to fit_data_and_texts or something similar.

[gleckler1]

@doutriaux1 Something like fit_data_and_texts would be good. In some cases a user will want to combine lots of postage stamp sized figures, but the real appeal of EzTemplate is to enable users to quickly put together an array of panels (~ < 6x3) and for that having data+text controlled together would be a very powerful and popular capability

[doutriaux1]

@gleckler1 will add. But ultimately you can't avoid having to do the final tweaks manually. For mass production I actually wouldn't recommend using EzTemplate. I think the way to go is to use EzTemplate first to get 99% of the way, thendo the last % manually, then save the templates and load them up at the beginning of mass production. @lee1043 I recommend you do this as well. It is faster to load a template from a json file, that using EzTemplate and move every attribute. It also definitely more memory efficient.