feat: HistStack for #169

[LovelyBuggies]

To solve #169.

[sourcery-ai]

Sourcery Code Quality Report

❌  Merging this PR will decrease code quality in the affected files by 0.05%.

Quality metrics	Before	After	Change
Complexity	0.67 ⭐	0.67 ⭐	0.00
Method Length	27.00 ⭐	28.00 ⭐	1.00 👎
Working memory	3.89 ⭐	3.89 ⭐	0.00
Quality	92.16% ⭐	92.11% ⭐	-0.05% 👎

Other metrics	Before	After	Change
Lines	38	40	2

Changed files	Quality Before	Quality After	Quality Change
src/hist/init.py	92.16% ⭐	92.11% ⭐	-0.05% 👎

Here are some functions in these files that still need a tune-up:

File	Function	Complexity	Length	Working Memory	Quality	Recommendation

Legend and Explanation

The emojis denote the absolute quality of the code:

• ⭐ excellent
• 🙂 good
• 😞 poor
• ⛔ very poor

The 👍 and 👎 indicate whether the quality has improved or gotten worse with this pull request.

---

Please see our documentation here for details on how these metrics are calculated.

We are actively working on this report - lots more documentation and extra metrics to come!

Help us improve this quality report!

[LovelyBuggies]

Maybe with more plotting options here. Any suggestions?

[henryiii]

h =  Hist.new.Reg(..., name="x").StrCat(["one", "two"] name="y").Double()

s = h.stack("y")
s = Stack({"one": h[:, "one"], "two":h[:, "two"]})

# Another way?
h1 = h[:, "one"]
h1.name = "one"
h2 = h[:, "two"]
h2.name = "two"
s = Stack(h1, h2)

h.plot() # -> shortcut for h.stack("y").plot()

Thought: Could we above a initial memory copy for stacking an existing histogram? May not be better, due to making copy every access?

self._data = h
self[x] # -> pull x from histogram _OR_ list/dict?

[LovelyBuggies]

We also need some tests for Stack(axes) and Hist.stack(name/idx)

Done.

[henryiii]

I think we should only support one arg, which is the axis to stack over. Multiple args would make a ND stack, which we don't want to support, at least for now.

The argument should be the axis to stack over, which should be expanded. For example:

stack_histograms = self[{ax:i}] for i in range(len(self.axis[ax]))]
return Stack(*stack_histograms)

Though I think we also need to keep track of the bin this came from too.

[LovelyBuggies]

What's the ax in this example?

[henryiii]

In my case, it was the integer axis number to expand. Though if it was a string axis name, since that is supported natively in both h[{...}] and h.axis[...], that should work too. I didn't expand the dict correctly above, though.

[henryiii]

Updated the example above - it now no longer "remembers" the cell name, though. Ideally, each histogram would also have at least "h.name" set to the original cell "name". Maybe:

for name, h in zip(self.axes[ax], stack_histograms)
    h.name = name

[henryiii]

If self.axes[i] is a StrCategory, which is most common, then h.name will be the category that this histogram came from. If it's an integer, then name is an integer - slightly weird, but not too bad. If it's a Regular/Variable, then "name" is a tuple with two floats, which is quite weird for something called "name". But I think it's a reasonable tradeoff - having the value is still useful.

[henryiii]

Quick summary of changes: axes are not stackable, only histograms. Added a little bit of container-like support so stack[0] gets the 0'th histogram in the stack. Added .name which gets set from the category (but currently unused by mplhep or us). Did a little parametrization of the tests (more could be done eventually). I'm reserving overlay= because I forgot what it was supposed to do. I also used stack.py to match Stack, and used self.__class__.__name__ in the repr.

[henryiii]

Thanks everyone!

[LovelyBuggies]

Added .name which gets set from the category (but currently unused by mplhep or us).

How to use .name? And h.stack() is all forbidden? There's no passed tests for .stack.

[henryiii]

I thought there were some. This is what it looks like:

h = hist.Hist.new.Reg(10,0,1, name="x").StrCategory(["good", "bad"], name="quality").Double()
# Fill with data tagged with quality="good" or "bad"
stack = h.stack("quality")
assert stack[0].name == "good"
assert stack[1].name == "bad"

Ideally, calling stack.plot(); plt.legend() would produce a legend that labels these with "good" and "bad", but that doesn't happen yet, might need to be added to mplhep. Unless there's a labels keyword or something, which I didn't check for.

Could you add a test like the one above, actually? I'm working on three other features.

[henryiii]

Oops, forgot, StrCat. Allowing both StrCat and StrCategory is allowed on the branch I'm working on - and accidentally passing the quick constructor to the Hist as an axis is allowed, too!

[LovelyBuggies]

FYI,

[henryiii]

This is why we write tests! 😳