Merge pull request #19156 from aschackmull/misc/stageoverlap

Misc: Add stage overlap script

Author: aschackmull

misc/scripts/stageoverlap.py

@@ -0,0 +1,81 @@
+#!/usr/bin/env python3
+
+import sys
+import os
+import re
+
+# read first argument
+if len(sys.argv) < 2:
+    print("Usage: stageoverlap.py <dil>")
+    sys.exit(1)
+
+dilfile = sys.argv[1]
+
+seen_stages = set()
+computed_predicates = {}
+stage_number = 0
+
+def process_stage(stage, cached):
+    global stage_number
+    stage_key = ' '.join(cached)
+    # skip repeated stages (in case we're looking at DIL for several queries, e.g. from a .qls)
+    if stage_key in seen_stages:
+        return
+    # don't count the query-stage as seen, since we don't want to skip those
+    if not '#select' in cached:
+        seen_stages.add(stage_key)
+    stage_number += 1
+    print('STAGE ' + str(stage_number) + ':')
+    print(str(len(cached)) + ' cached predicate(s)')
+    print(' '.join(cached))
+    for predicate in stage:
+        # strip trailing characters matching the regex '#[bf]+', i.e. disregard magic
+        predicate = re.sub('#[bf]+$', '', predicate)
+        # TODO: maybe also strip the hash?
+        # predicate = re.sub('#[a-f0-9]+$', '', predicate)
+        if predicate in computed_predicates.keys():
+            # skip db-relations and some generated predicates
+            if predicate.startswith('@') or predicate.startswith('project#'):
+                continue
+            prior_stage = computed_predicates[predicate]
+            print('Recompute from ' + str(prior_stage) + ': ' + predicate)
+        else:
+            computed_predicates[predicate] = stage_number
+    print()
+
+with open(dilfile, 'r') as f:
+    stage = []
+    cached = []
+    query = False
+    for line in f:
+        # skip lines starting with a space, i.e. predicate bodies
+        if line.startswith(' '): continue
+        # get the part of the line containing no spaces occuring before the first '('
+        # this is the predicate name
+        parenpos = line.find('(')
+        if parenpos != -1:
+            start = line.rfind(' ', 0, parenpos)
+            predicate = line[start+1:parenpos]
+            if predicate.startswith('`'):
+                # remove the leading and trailing backticks
+                predicate = predicate[1:-1]
+            stage.append(predicate)
+            continue
+        # query predicates, aka cached predicates, are written either as
+        # 'query <predicatename> = ...' on one line, or split across 2+ lines
+        if line.startswith('query '):
+            predicate = line.split(' ')[1]
+            cached.append(predicate)
+            continue
+        if line == 'query\n':
+            query = True
+            continue
+        if query:
+            predicate = line.split(' ')[0]
+            cached.append(predicate)
+            query = False
+            continue
+        if line == '/* ---------- END STAGE ---------- */\n':
+            process_stage(stage, cached)
+            stage = []
+            cached = []