Add flexible plots legends to plots tree

[mattseddon]

2/2 main <- #2403 <- this

Related to #1757

This PR adds shape and stroke dash information into the plots tree.

Screenshots

Note: The hacky way that we add shape + detail to the plot template causes the on-hover behaviour to break. I will change the approach in the next PR.

[mattseddon]

In order to test set MIN_CLI_VERSION = '2.0.0' and use the same repository from #2403. In order to get shapes, I updated the script to the following:

#!/bin/bash

set -exu

wsp=test_wspace_1
rep=test_repo

rm -rf $wsp && mkdir $wsp && pushd $wsp
main=$(pwd)

mkdir $rep && pushd $rep

git init
dvc init

mkdir subdir && pushd subdir

git add -A
git commit -am "initial"

echo "some_param: 0.1" > params.yaml


echo "
import yaml
import json
import random

def get_params():
    with open('params.yaml', 'r') as fd:
        return yaml.safe_load(fd)
p = get_params()['some_param']


def prepare_cm(num_classes=3,num_samples=3):
    result = []
    for i in range(num_classes):
        for j in range(num_samples):
            if random.random() <= p:
                result.append({'actual': i, 'predicted': i})
            else:
                result.append({'actual': i, 'predicted': random.choice(list(range(num_samples)))})
    return result

res1=[]
res2=[]
for i in range(1, 11):
    res1.append({'x': i, 'y': i*p})
    res2.append({'x': i, 'z': i*p*2, 'y': i*p*random.random()})

with open('res1.json', 'w') as fd:
    json.dump(res1, fd)

with open('res2.json', 'w') as fd:
    json.dump(res2, fd)

with open('metric1.json', 'w') as fd:
    json.dump(res1[-1],fd)

with open('metric2.json', 'w') as fd:
    json.dump(res2[-1],fd)

with open('confusion_matrix.json', 'w') as fd:
    json.dump(prepare_cm(), fd)

import shutil
shutil.copyfile('confusion_matrix.json', 'confusion_matrix2.json')
" > train.py

dvc run -d train.py -p some_param \
--plots-no-cache confusion_matrix.json \
--plots-no-cache res1.json \
-O res2.json -M metric1.json -M metric2.json \
-O confusion_matrix2.json -n train "python train.py"

echo "plots:
  res2.json:
    template: simple
  combined:
    x: x
    y:
      res1.json: y
      res2.json: [y,z]
  confusion_matrix2.json:
    template: confusion
    x: actual
    y: predicted" >> dvc.yaml
 

dvc plots modify res1.json --template scatter -x x -y y
dvc plots modify confusion_matrix.json --template confusion -x actual -y predicted

git add -A
git commit -am "initial run"

echo "some_param: 0.3
" > params.yaml

dvc exp run -n "p0.9"

echo "some_param: 0.9
" > params.yaml

dvc repro

[sroy3]

Should these be set where it's enabled only? (Line 311)

[mattseddon]

I am going to have to change the approach slightly now that we don't know exactly where the encoding update has been made.

[mattseddon]

I will call out the reasoning for leaving in the follow-up.

[sroy3]

What changed here? Is that only a lint/prettier config?

[mattseddon]

yeah, I can revert

[sroy3]

That's fine, I was just wondering if there is something I was missing. Better keep it that way IMO

[qlty-cloud-legacy]

Code Climate has analyzed commit e10c8d6 and detected 2 issues on this pull request.

Here's the issue category breakdown:

Category	Count
Duplication	2

The test coverage on the diff in this pull request is 94.9% (85% is the threshold).

This pull request will bring the total coverage in the repository to 96.9% (0.0% change).

View more on Code Climate.