A script to summarize gas differences from isoltest for PRs.

Author: hrkrshnn

scripts/gas_diff_stats.py

@@ -0,0 +1,139 @@
+"""A script to collect gas statistics and print it.
+
+Useful to summarize gas differences to semantic tests for a PR / branch.
+
+Dependencies: Parsec (https://pypi.org/project/parsec/) and Tabulate
+(https://pypi.org/project/tabulate/)
+
+  pip install parsec tabulate
+
+Run from root project dir.
+
+  python3 scripts/gas_diff_stats.py
+
+Note that the changes to semantic tests have to be committed.
+
+Assumes that there is a remote named ``origin`` pointing to the Solidity github
+repository. The changes are compared against ``origin/develop``.
+
+"""
+import subprocess
+from pathlib import Path
+from enum import Enum
+from parsec import *
+from tabulate import tabulate
+
+class Kind(Enum):
+    IrOptimized = 1
+    Legacy = 2
+    LegacyOptimized = 3
+
+class Diff(Enum):
+    Minus = 1
+    Plus = 2
+
+minus = string("-").result(Diff.Minus)
+plus = string("+").result(Diff.Plus)
+
+space = string(" ")
+comment = string("//")
+colon = string(":")
+
+gas_ir_optimized = string("gas irOptimized").result(Kind.IrOptimized)
+gas_legacy_optimized = string("gas legacyOptimized").result(Kind.LegacyOptimized)
+gas_legacy = string("gas legacy").result(Kind.Legacy)
+
+def number() -> int:
+    """Parse number."""
+    return regex(r"([0-9]*)").parsecmap(int)
+
+@generate
+def diff_string() -> (Kind, Diff, int):
+    """Usage: diff_string.parse(string)
+
+    Example string:
+
+    -// gas irOptimized: 138070
+
+    """
+    diff_kind = yield (minus | plus)
+    yield comment
+    yield space
+    codegen_kind = yield (gas_ir_optimized ^ gas_legacy_optimized ^ gas_legacy)
+    yield colon
+    yield space
+    val = yield number()
+    return (diff_kind, codegen_kind, val)
+
+def collect_statistics(lines) -> (int, int, int, int, int, int):
+    """Returns
+
+    (old_ir_optimized, old_legacy_optimized, old_legacy, new_ir_optimized,
+    new_legacy_optimized, new_legacy)
+
+    All the values in the same file (in the diff) are summed up.
+
+    """
+    if not lines:
+        raise Exception("Empty list")
+
+    def try_parse(line):
+        try:
+            return diff_string.parse(line)
+        except ParseError:
+            pass
+
+    out = [parsed for line in lines if (parsed := try_parse(line)) is not None]
+    diff_kinds = [Diff.Minus, Diff.Plus]
+    codegen_kinds = [Kind.IrOptimized, Kind.LegacyOptimized, Kind.Legacy]
+    return tuple([
+        sum([
+            val
+            for (diff_kind, codegen_kind, val) in out
+            if diff_kind == _diff_kind and codegen_kind == _codegen_kind
+        ])
+        for _diff_kind in diff_kinds
+        for _codegen_kind in codegen_kinds
+    ])
+
+def semantictest_statistics():
+    """Prints the tabulated statistics that can be pasted in github."""
+    def try_parse_git_diff(fname):
+        try:
+            diff_output = subprocess.check_output(
+                "git diff --unified=0 origin/develop HEAD " + fname,
+                shell=True,
+                universal_newlines=True
+            ).splitlines()
+            if diff_output:
+                return collect_statistics(diff_output)
+        except subprocess.CalledProcessError as e:
+            print("Error in the git diff:")
+            print(e.output)
+        return None
+    def stat(old, new):
+        return ((new - old) / old) * 100  if old else 0
+
+    table = []
+
+    for path in Path("test/libsolidity/semanticTests").rglob("*.sol"):
+        fname = path.as_posix()
+        parsed = try_parse_git_diff(fname)
+        if parsed is None:
+            continue
+        ir_optimized = stat(parsed[0], parsed[3])
+        legacy_optimized = stat(parsed[1], parsed[4])
+        legacy = stat(parsed[2], parsed[5])
+        fname = fname.split('/', 3)[-1]
+        table += [map(str, [fname, ir_optimized, legacy_optimized, legacy])]
+
+    if table:
+        print("<details><summary>Click for a table of gas differences</summary>\n")
+        table_header = ["File name", "IR-optimized (%)", "Legacy-Optimized (%)", "Legacy (%)"]
+        print(tabulate(table, headers=table_header, tablefmt="github"))
+        print("</details>")
+    else:
+        print("No differences found.")
+
+if __name__ == "__main__":
+    semantictest_statistics()