ghc_timings: Use pandas

default.nix

@@ -38,6 +38,8 @@ let
           --prefix PATH : ${python3.withPackages (pkgs: with pkgs; [ matplotlib pandas tabulate ])}/bin
         makeWrapper ${./sample_proc.py} $out/bin/sample-proc \
           --prefix PATH : ${python3.withPackages (pkgs: with pkgs; [ matplotlib pandas tabulate ])}/bin
+        makeWrapper ${./ghc_timings.py} $out/bin/ghc-timings \
+          --prefix PATH : ${python3.withPackages (pkgs: with pkgs; [ matplotlib pandas tabulate ])}/bin
       '';
     };
   gdb = import ./gdb { inherit nixpkgs; };

ghc_timings.py

@@ -4,6 +4,7 @@
 import re
 import sys
 import typing
+import pandas as pd
 from typing import Iterator
 
 class Timing(typing.NamedTuple):
@@ -12,7 +13,7 @@ class Timing(typing.NamedTuple):
     time: float
     allocd: float
 
-def parse_timings(s: str) -> Iterator[Timing]:
+def parse_timings_raw(s: str) -> Iterator[Timing]:
     for m in re.finditer(r'!!! ([\w \d]+) \[([\w\d_\.]+)\]: finished in (\d+\.\d+) milliseconds, allocated (\d+\.\d+) megabytes', s):
         phase = m.group(1)
         module = m.group(2)
@@ -20,12 +21,16 @@ def parse_timings(s: str) -> Iterator[Timing]:
         allocd = float(m.group(4))
         yield Timing(phase, module, time, allocd)
 
+def parse_timings(s: str) -> pd.DataFrame:
+    xs = list(parse_timings_raw(s))
+    df = pd.DataFrame(xs, columns=['phase', 'module', 'time', 'allocd'])
+    df['n'] = df.index
+    df.sort_values(by=['module', 'n'], inplace=True)
+    return df.drop(columns='n')
+
 def main() -> None:
     s = sys.stdin.read()
-    print('| module | phase | time (ms) | alloc (MB) |')
-    print('| ------ | ----- | --------- | ---------- |')
-    for t in parse_timings(s):
-        print('| {module} | {phase} | {time} | {allocd} |'.format(**t))
+    print(parse_timings(s).to_markdown())
 
 if __name__ == '__main__':
     main()