Replace plotting with polars

requirements.txt

@@ -1,7 +1,6 @@
 jinja2
 matplotlib
 numpy
-pandas
 polars
 scipy
 sympy>=1.9

tools/mapping-tester/plotconv.py

@@ -5,8 +5,7 @@
 
 import matplotlib.pyplot as plt
 import numpy as np
-import pandas
-from pandas.api.types import is_numeric_dtype
+import polars as pl
 
 
 def parseArguments(args):
@@ -29,10 +28,6 @@ def parseArguments(args):
     return parser.parse_args(args)
 
 
-def lavg(l):
-    return math.exp(sum(map(math.log, l)) / len(l))
-
-
 # seaborn.color_palette("colorblind", 10).as_hex()
 style_colours = [
     "#0173b2",
@@ -50,148 +45,36 @@ def lavg(l):
 styles = [(c, m) for m in style_markers for c in style_colours]
 
 
-def plotConv(ax, df, yname):
-    xmin = df["mesh A"].min()
-    xmax = df["mesh A"].max()
-    ymin = df[yname].min()
-    ymax = df[yname].max()
-
-    print(xmin, xmax)
-    print(ymin, ymax)
-
-
-def plotError(df, prefix):
-    yname = "relative-l2"
-    if yname not in df.columns:
-        print(f"Skipping {yname} as not found in dataset.")
-        return
-    fig, ax = plt.subplots(sharex=True, sharey=True)
-    series = df.groupby("mapping")
-    for grouped, style in zip(series, styles):
-        name, group = grouped
-        if group[yname].max() == 0:
-            print(f"Dropping {yname}-series {name} as all 0")
-            continue
-        color, marker = style
-        group.plot(
-            ax=ax,
-            logy=True,
-            logx=is_numeric_dtype(df["mesh A"]),
-            x="mesh A",
-            y=yname,
-            label=name,
-            marker=marker,
-            color=color,
-        )
-    ax.set_xlabel("edge length(h) of mesh A")
-    ax.set_ylabel("relative-l2 error mapping to mesh B")
-
-    plotConv(ax, df, yname)
-
-    plt.gca().invert_xaxis()
-    plt.grid()
-    plt.savefig(prefix + "-error.pdf")
-
-
-def plotMemory(df, prefix):
-    yname = "peakMemB"
+def plot(df: pl.DataFrame, yname: str, ylabel: str, filename: str):
     if yname not in df.columns:
         print(f"Skipping {yname} as not found in dataset.")
         return
-    fig, ax = plt.subplots(sharex=True, sharey=True)
-    series = df.groupby("mapping")
-    for grouped, style in zip(series, styles):
-        name, group = grouped
-        if group[yname].max() == 0:
-            print(f"Dropping {yname}-series {name} as all 0")
-            continue
-        color, marker = style
-        group.plot(
-            ax=ax,
-            logy=True,
-            logx=is_numeric_dtype(df["mesh A"]),
-            x="mesh A",
-            y=yname,
-            label=name,
-            marker=marker,
-            color=color,
-        )
-    ax.set_xlabel("edge length(h) of mesh A")
-    ax.set_ylabel("peak memory of participant B [Kbytes]")
 
-    # plotConv(ax, df, yname)
-
-    plt.gca().invert_xaxis()
-    plt.grid()
-    plt.savefig(prefix + "-peakMemB.pdf")
-
-
-def plotComputeMappingTime(df, prefix):
-    yname = "computeMappingTime"
-    if yname not in df.columns:
-        print(f"Skipping {yname} as not found in dataset.")
-        return
     fig, ax = plt.subplots(sharex=True, sharey=True)
-    series = df.groupby("mapping")
-    for grouped, style in zip(series, styles):
-        name, group = grouped
-        if group[yname].max() == 0:
-            print(f"Dropping {yname}-series {name} as all 0")
-            continue
-        color, marker = style
-        group.plot(
-            ax=ax,
-            logy=True,
-            logx=is_numeric_dtype(df["mesh A"]),
-            x="mesh A",
-            y=yname,
-            label=name,
-            marker=marker,
-            color=color,
-        )
-
     ax.set_xlabel("edge length(h) of mesh A")
-    ax.set_ylabel("time to compute mapping [us]")
+    ax.set_ylabel(ylabel)
+    ax.set_yscale("log")
+    if df["mesh A"].dtype.is_numeric():
+        ax.set_xscale("log")
 
-    # plotConv(ax, df, yname)
-
-    plt.gca().invert_xaxis()
-    plt.grid()
-    plt.savefig(prefix + "-computet.pdf")
-
-
-def plotMapDataTime(df, prefix):
-    yname = "mapDataTime"
-    if yname not in df.columns:
-        print(f"Skipping {yname} as not found in dataset.")
-        return
-    fig, ax = plt.subplots(sharex=True, sharey=True)
-    series = df.groupby("mapping")
+    series = df.group_by("mapping")
     for grouped, style in zip(series, styles):
         name, group = grouped
         if group[yname].max() == 0:
             print(f"Dropping {yname}-series {name} as all 0")
             continue
         color, marker = style
-        group.plot(
-            ax=ax,
-            logy=True,
-            logx=is_numeric_dtype(df["mesh A"]),
-            x="mesh A",
-            y=yname,
+        ax.plot(
+            group["mesh A"],
+            group[yname],
             label=name,
             marker=marker,
             color=color,
         )
 
-    ax.set_xlabel("edge length(h) of mesh A")
-    ax.set_ylabel("time to map Data [us]")
-
-    # plotConv(ax, df, yname)
-
     plt.gca().invert_xaxis()
     plt.grid()
-    plt.savefig(prefix + "-mapt.pdf")
+    plt.savefig(filename + ".pdf")
 
 
 def main(argv):
@@ -201,16 +84,40 @@ def main(argv):
     plt.rcParams["figure.figsize"] = "8, 8"
     plt.rcParams["figure.autolayout"] = "true"
 
-    df = pandas.read_csv(args.file)
+    df = pl.read_csv(args.file).sort("mesh A")
     toMeshes = df["mesh B"].unique()
     assert (
         len(toMeshes) == 1
     ), f"There are {len(toMeshes)} to-meshes but only 1 is allowed. Fix your dataset!"
-    df.sort_values("mesh A", inplace=True)
-    plotError(df, args.prefix)
-    plotMemory(df, args.prefix)
-    plotMapDataTime(df, args.prefix)
-    plotComputeMappingTime(df, args.prefix)
+
+    if not df["mesh A"].dtype.is_numeric():
+        print("Note: 'mesh A' isn't numeric. The x-axis will not use log scaling.")
+
+    plot(
+        df,
+        yname="relative-l2",
+        ylabel="relative-l2 error mapping to mesh B",
+        filename=f"{args.prefix}-error",
+    )
+    plot(
+        df,
+        yname="peakMemB",
+        ylabel="peak memory of participant B [Kbytes]",
+        filename=f"{args.prefix}-peakMemB",
+    )
+    plot(
+        df,
+        yname="computeMappingTime",
+        ylabel="time to compute mapping [us]",
+        filename=f"{args.prefix}-computet",
+    )
+    plot(
+        df,
+        yname="mapDataTime",
+        ylabel="time to map Data [us]",
+        filename=f"{args.prefix}-mapt",
+    )
+
     return 0