|
5 | 5 | import pandas as pd |
6 | 6 | import seaborn as sns |
7 | 7 |
|
8 | | -json_path = os.environ.get("BENCHMARK_DATA", "benchmark_data") |
| 8 | +data_path = os.environ.get("BENCHMARK_DATA", "benchmark_data") |
9 | 9 | plot_path = os.environ.get("BENCHMARK_PLOT", "benchmark_plot.png") |
10 | | - |
11 | 10 | xy = os.environ.get("XY", "unknown") |
12 | 11 |
|
13 | | -# s3+hdf5, s3+tiff, s3+zarr, remote+hdf5, remote+… so I’d color by tiff/hdf5/zarr |
14 | | - |
15 | | -print("base_path", json_path) |
16 | | - |
17 | | - |
18 | | -three_col = [] |
19 | | -two_col = [] |
20 | | - |
21 | | -test_repeats = int(os.getenv("TEST_REPEATS", "1")) |
22 | | -json_files = ["%s_benchmark_data.json" % r for r in range(test_repeats)] |
23 | | -print("json_files", json_files) |
24 | | - |
25 | | -for root, dirs, files in os.walk(json_path): |
26 | | - for file_name in files: |
27 | | - if file_name in json_files: |
28 | | - path = os.path.join(root, file_name) |
29 | | - print("json path", path) |
30 | | - with open(path) as json_file: |
31 | | - benchmarks = json.load(json_file)["benchmarks"] |
| 12 | +csv = pd.read_csv(f"{data_path}/0_benchmark_data.csv") |
32 | 13 |
|
33 | | - for bm in benchmarks: |
34 | | - m = re.match(r"test_(1_byte|\w+)_(\w+)\[(\w+)\]", bm["name"]) |
35 | | - if not m: |
36 | | - raise Exception(bm["name"]) |
37 | | - typ = m.group(1).replace("1_byte", "overhead") |
38 | | - src = m.group(3) |
39 | | - |
40 | | - if typ == "overhead" and src == "local": |
41 | | - # 10e-5 skews the view. |
42 | | - continue |
43 | | - |
44 | | - if test_repeats == 1: |
45 | | - # Ran tests once: plot every data point |
46 | | - vals = bm["stats"]["data"] |
47 | | - for run, val in enumerate(vals): |
48 | | - three_col.append( |
49 | | - {"type": typ, "source": src, "seconds": val} |
50 | | - ) |
51 | | - two_col.append({"name": f"{typ}-{src}", "seconds": val}) |
52 | | - else: |
53 | | - # Repeats: take mean value from each |
54 | | - val = bm["stats"]["mean"] |
55 | | - three_col.append({"type": typ, "source": src, "seconds": val}) |
56 | | - two_col.append({"name": f"{typ}-{src}", "seconds": val}) |
57 | | - |
58 | | -df3 = pd.DataFrame.from_dict(three_col) |
59 | | -df2 = pd.DataFrame.from_dict(two_col) |
60 | | - |
61 | | -types = ("overhead", "zarr", "tiff", "hdf5") |
| 14 | +types = ("Overhead", "Zarr", "TIFF", "HDF5") |
62 | 15 | sources = ("local", "http", "s3") |
63 | 16 | orders = {"type": types, "source": sources} |
64 | 17 |
|
65 | 18 | pal_points = "colorblind" |
66 | 19 | pal_violins = "pastel" |
67 | 20 |
|
68 | 21 | g = sns.FacetGrid( |
69 | | - df3, col="source", col_order=sources, sharey=False, height=5, aspect=0.6, |
| 22 | + csv, |
| 23 | + col="source", |
| 24 | + col_order=sources, |
| 25 | + sharey=False, |
| 26 | + height=5, |
| 27 | + aspect=0.6, |
70 | 28 | ) |
71 | 29 |
|
72 | 30 | g = g.map( |
|
83 | 41 |
|
84 | 42 | g = g.map( |
85 | 43 | sns.stripplot, |
86 | | - "type", |
| 44 | + "type", # was type |
87 | 45 | "seconds", |
88 | 46 | dodge=True, |
89 | 47 | order=types, |
|
93 | 51 | ) |
94 | 52 |
|
95 | 53 | g.despine(left=True) |
96 | | -g.set(yscale="log", ylim=(0.0009, 1)) |
| 54 | +g.set(yscale ='log', ylim=(0.0009, 1)) |
97 | 55 |
|
98 | 56 | # Set axis labels & ticks # |
99 | 57 | for ax in g.fig.get_axes(): |
|
105 | 63 | # Remove outline of violins |
106 | 64 | col.set_edgecolor("white") |
107 | 65 |
|
| 66 | +# Only show on far left plot |
108 | 67 | g.fig.get_axes()[0].set_ylabel("Seconds") |
109 | 68 | g.fig.get_axes()[0].spines["left"].set_visible(True) |
110 | 69 |
|
| 70 | +# Add annotations |
| 71 | +g.fig.get_axes()[0].text(0.0001, 0.001, "off-\nscale") |
| 72 | + |
111 | 73 | g.savefig(plot_path, dpi=600) |
0 commit comments