b712b918c9fdf2aac65b5f6ceb076e4af584231c braney Tue Sep 1 12:29:36 2026 -0700 trackDbConditions: scan the whole library, and check the scanner against known cases, refs #37908 Four refinements, one of which fixes wrong output rather than noisy output. The scanned file list was hand-kept, and it had silently missed netCart.c, chainCart.c, pgSnp.c, hgMaf.c and a dozen more that read track settings on the drawing path. An unscanned read site is worse than an unclassified one: the every-path analysis was claiming a condition holds at every read of a setting while never having seen one of the reads. Three settings were carrying false claims because of it, barChartBars, barChartCategoryUrl and bigDataUrl. So scan hg/lib and hg/cgilib whole and let reachability decide which side of the browser each read belongs to. Coverage goes from 325 settings to 347. An early return is treated as a precondition only near the top of a function. The same shape four hundred lines down is sound but says nothing about the setting, and it was how the jsonp output check at the tail of doTrackForm came to look like a condition on filterBy. A negated disjunction is a conjunction, so NOT (A || B) now splits into NOT A and NOT B. The squishyPack guard was one unsplittable string that was neither a visibility condition nor a coverage one; it is now correctly both. --self-test checks the harvest against ten cases read out of the C by hand. Every one of them broke at least once while this was being built, usually silently, so they are checked rather than trusted, and --check runs them first and refuses to report anything if the scanner itself has moved. It earned its place immediately by catching two misclassifications in the same commit that added it. Also caches the harvest in a temp file keyed on the newest source mtime, since scanning takes forty seconds and reading the output takes several runs. Warm runs are now instant. diff --git src/hg/utils/trackDbConditions/trackDbConditions.py src/hg/utils/trackDbConditions/trackDbConditions.py index d5691c18ab5..0c50921b978 100755 --- src/hg/utils/trackDbConditions/trackDbConditions.py +++ src/hg/utils/trackDbConditions/trackDbConditions.py @@ -23,34 +23,37 @@ trackDbConditions.py --check # for a cron; see below A condition is reported as *always* when it holds at every place the browser reads that setting, and as *sometimes* when it holds at one place and not another. Only *always* is a claim about the setting; *sometimes* is a pointer to a call site worth reading. --check is the cron mode. It fails when a setting the documentation describes gains its first always-condition, or loses its last one, because either way the documentation now says something different from the code. The accepted state lives in conditionBaseline.txt beside this file; --update-baseline accepts. """ import argparse import collections +import getpass +import glob import json import os import re import sys +import tempfile sys.path.insert(0, os.path.dirname(os.path.abspath(__file__))) import harvestConditions as hc # noqa: E402 BASELINE = os.path.join(os.path.dirname(os.path.abspath(__file__)), "conditionBaseline.txt") SETTINGS_JSON = "hg/htdocs/goldenPath/help/trackDb/trackDbSettings.json" # What kind of thing has to be true. Order matters: the first match wins, and # the earlier entries are the specific ones. KINDS = [ ("coverage", "density coverage mode", r"checkIfWiggling|winTooBigDoWiggle|limitWiggle|\bdoWiggle\b|setupForWiggle"), ("snake", "snake mode", r"\bdoSnake\b|snakeMode"), ("multiRegion", "multi-region mode", @@ -88,88 +91,139 @@ checkIfWiggling() is doWiggle, hicUiFetchAutoScale() is autoScale. A condition that calls one of these is testing a setting, and saying so is more use than printing the function name. The test is deliberately strict: the function must hold exactly one track-setting read, and that read must not itself sit behind a condition, or a large function with one incidental read would be mistaken for an accessor. """ byFunc = collections.defaultdict(list) for read in reads: if read["tdb"] and read["func"]: byFunc[read["func"]].append(read) return {func: rs[0]["name"] for func, rs in byFunc.items() if len(rs) == 1 and not rs[0]["conds"]} -def classify(cond, accessors=None): +def classify(cond, accessors=None, settingNames=None): """Which kind of condition this is, and the settings it names if any.""" text = cond["text"] named = list(cond.get("derivedFrom") or []) for m in re.finditer(r'"([A-Za-z][A-Za-z0-9_.]*)"', text): if READER_CALL.search(text): named.append(m.group(1)) for m in re.finditer(r"([A-Za-z_][A-Za-z0-9_]*)\s*\(", text): if accessors and m.group(1) in accessors: named.append(accessors[m.group(1)]) if named: return OTHER_SETTING, sorted(set(named)) for kind, _, pattern in KINDS: if re.search(pattern, text): return kind, [] + # Last, and weakest: a bare variable named after a setting is usually + # holding it, which is how drawMode arrives from hicUiFetchDrawMode. It runs + # after the patterns above on purpose. track->visibility is the runtime + # field, not the visibility setting, and reading it the other way round put + # the squishyPack guard in the wrong bucket. Field accesses are skipped for + # the same reason, and short names because type, name and color are settings + # and also ordinary words. + bare = re.findall(r"(?<![\w>.])([A-Za-z_][A-Za-z0-9_]*)", text) + hits = sorted({ident for ident in bare + if settingNames and ident in settingNames and len(ident) >= 5}) + if hits: + return OTHER_SETTING, hits return UNCLASSIFIED, [] KIND_TEXT = dict([(k, t) for k, t, _ in KINDS] + [(OTHER_SETTING, OTHER_SETTING_TEXT), (UNCLASSIFIED, "something else")]) KIND_ORDER = [k for k, _, _ in KINDS] + [OTHER_SETTING, UNCLASSIFIED] def condId(cond): return re.sub(r"\s+", "", cond["text"]) -def build(kentSrc=None, cache=None): +def defaultCache(kentSrc): + """Where to keep the harvest between runs. + + Scanning the tree takes about forty seconds, which is fine once and tiresome + four times in a row while reading the output. The cache is keyed on the + newest source file, so editing any scanned file rebuilds it, and it lives in + the temp directory rather than the tree so it cannot dirty a checkout. + """ + tag = re.sub(r"\W+", "_", os.path.abspath(kentSrc)).strip("_")[-60:] + return os.path.join(tempfile.gettempdir(), + "trackDbConditions-%s-%s.json" % (getpass.getuser(), tag)) + + +def sourceStamp(kentSrc): + """The newest modification time across everything the harvest reads.""" + newest = 0.0 + patterns = set(hc.SHARED) + for pats in hc.SCOPES.values(): + patterns.update(pats) + for pattern in patterns: + for path in glob.glob(os.path.join(kentSrc, pattern)): + try: + newest = max(newest, os.path.getmtime(path)) + except OSError: + pass + return round(newest, 3) + + +def build(kentSrc=None, cache=None, refresh=False): """Harvest, classify, and fold the read sites together per setting and scope.""" kentSrc = kentSrc or hc.os.environ.get("KENT_SRC") or os.path.abspath( os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "..", "..")) - if cache and os.path.exists(cache): + cache = cache or defaultCache(kentSrc) + stamp = sourceStamp(kentSrc) + reads = None + if not refresh and os.path.exists(cache): + try: with open(cache) as f: - reads = json.load(f) - else: + held = json.load(f) + if held.get("stamp") == stamp: + reads = held["reads"] + except (ValueError, KeyError, OSError): + reads = None + if reads is None: reads = hc.harvest(kentSrc) - if cache: + try: with open(cache, "w") as f: - json.dump(reads, f) + json.dump({"stamp": stamp, "reads": reads}, f) + except OSError: + pass accessors = accessorMap(reads) + settingNames = {r["name"] for r in reads if r["tdb"]} settings = {} for read in reads: if not read["tdb"]: continue # a page cart variable, not a track setting key = (read["name"], read["scope"]) entry = settings.setdefault(key, {"name": read["name"], "scope": read["scope"], "sites": [], "unknown": False}) conds = [] for cond in hc.realConds(read): - kind, named = classify(cond, accessors) + kind, named = classify(cond, accessors, settingNames) conds.append({"text": cond["text"], "kind": kind, "names": named, "file": cond["file"], "line": cond["line"]}) used = [] for cond in read.get("useConds", []): if hc.isNoise(cond): continue - kind, named = classify(cond, accessors) + kind, named = classify(cond, accessors, settingNames) used.append({"text": cond["text"], "kind": kind, "names": named, "file": cond["file"], "line": cond["line"]}) entry["sites"].append({"file": read["file"], "line": read["line"], "func": read["func"], "reader": read["reader"], "conds": conds, "used": used}) if read["callerUnknown"]: entry["unknown"] = True for entry in settings.values(): perSite = [{condId(c): c for c in s["conds"]} for s in entry["sites"]] always, sometimes = {}, {} if perSite: common = set(perSite[0]) for one in perSite[1:]: common &= set(one) @@ -231,73 +285,137 @@ groups.append(("sometimes", entry["sometimes"])) elif entry["sometimes"]: print(" (%d more conditions hold at some of its %d read sites; --verbose to see)" % (len(entry["sometimes"]), len(entry["sites"]))) for label, conds in groups: for cond in conds: names = (" -> " + ", ".join(cond["names"])) if cond["names"] else "" print(" %-9s %-12s %s%s" % (label, cond["kind"], cond["text"][:88], names)) if entry["unknown"]: print(" (some call paths could not be followed, so this is a floor)") if verbose: for site in entry["sites"]: print(" %s:%d %s()" % (site["file"], site["line"], site["func"])) +# Cases read out of the C by hand. Every one of them broke at least once while +# this was being built, usually silently, so they are checked rather than +# trusted. Each is (scope, setting, where, kind, text that must appear); a +# leading "!" on the text means it must NOT appear anywhere in that setting. +SELF_TEST = [ + ("render", "bamColorTag", "whenUsed", "otherSetting", "bamColorMode"), + ("render", "pairSearchRange", "whenUsed", "otherSetting", "pairEndsByName"), + ("render", "frames", "always", "visibility", "tvFull"), + ("render", "frames", "always", "zoom", "zoomedToBaseLevel"), + ("render", "squishyPackPoint", "always", "visibility", "tvPack"), + ("render", "hicArcLimit", "always", "otherSetting", "drawMode"), + ("render", "hapClusterHeight", "always", "coverage", "setupForWiggle"), + ("render", "hideEmptySubtracks", "always", "container", "tdbIsComposite"), + ("config", "minGrayLevel", "always", "otherSetting", "scoreMin"), + # the output-format check at the tail of doTrackForm is not a condition on + # a track setting, however sound the control flow makes it + ("render", "filterBy", "any", None, "!jsonp"), +] + + +def selfTest(entries): + """Check the harvest against cases confirmed by reading the code.""" + byKey = {(e["scope"], e["name"]): e for e in entries} + bad = 0 + for scope, name, where, kind, want in SELF_TEST: + entry = byKey.get((scope, name)) + if entry is None: + print("FAIL %s %s: not read at all in that scope" % (scope, name)) + bad += 1 + continue + if where == "any": + conds = entry["always"] + entry["whenUsed"] + entry["sometimes"] + else: + conds = entry["always" if where == "always" else "whenUsed"] + if want.startswith("!"): + hits = [c for c in conds if want[1:] in c["text"] or want[1:] in c["names"]] + if hits: + print("FAIL %s %s: should not mention %s, but %s does" + % (scope, name, want[1:], hits[0]["text"][:60])) + bad += 1 + continue + hits = [c for c in conds + if (kind is None or c["kind"] == kind) + and (want in c["text"] or want in c["names"])] + if not hits: + print("FAIL %s %s [%s]: no %s condition mentioning %s" + % (scope, name, where, kind, want)) + for c in conds: + print(" had: %-12s %s" % (c["kind"], c["text"][:70])) + bad += 1 + print("self test: %d of %d cases pass" % (len(SELF_TEST) - bad, len(SELF_TEST))) + return bad == 0 + + def main(): parser = argparse.ArgumentParser( description=__doc__.split("\n\n")[0], formatter_class=argparse.RawDescriptionHelpFormatter) parser.add_argument("--list", action="store_true", help="every setting with a condition") parser.add_argument("--setting", help="just this setting") parser.add_argument("--kind", help="just this kind of condition (%s)" % ", ".join(KIND_ORDER)) parser.add_argument("--scope", default="render", choices=list(hc.SCOPES) + ["both"]) parser.add_argument("--documented", action="store_true", help="only settings trackDbLibrary.shtml describes") parser.add_argument("--verbose", action="store_true", help="show the sometimes-conditions and the call sites too") parser.add_argument("--surprising", action="store_true", help="only settings whose condition is not just the track type") parser.add_argument("--json", help="write everything here") parser.add_argument("--check", action="store_true", help="cron mode, see the module docstring") + parser.add_argument("--self-test", action="store_true", + help="check the harvest against cases confirmed by hand") parser.add_argument("--update-baseline", action="store_true", help="accept the current set of conditioned settings") - parser.add_argument("--cache", help="reuse a harvest written here") + parser.add_argument("--cache", help="keep the harvest here instead of the default temp file") + parser.add_argument("--refresh", action="store_true", help="rescan even if the cache is current") args = parser.parse_args() kentSrc = os.environ.get("KENT_SRC") or os.path.abspath( os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "..", "..")) - entries = build(kentSrc, args.cache) + entries = build(kentSrc, args.cache, args.refresh) docs = documented(kentSrc) if args.json: with open(args.json, "w") as f: json.dump({"settings": entries, "kinds": KIND_TEXT}, f, indent=1) print("wrote %s" % args.json) + if args.self_test: + sys.exit(0 if selfTest(entries) else 1) + if args.update_baseline: names = sorted(conditionedNames(entries, docs)) with open(BASELINE, "w") as f: f.write("# Documented settings whose reads all sit behind a condition.\n" "# Accepted state for trackDbConditions.py --check. One scope:name per line.\n") for name in names: f.write(name + "\n") print("wrote %s (%d settings)" % (BASELINE, len(names))) return if args.check: + if not selfTest(entries): + print("the scanner itself is not reporting what it used to; " + "fix that before reading anything below") + sys.exit(1) now = conditionedNames(entries, docs) was = baselineNames() problems = 0 for name in sorted(now - was): print("newly conditional, and the docs do not say so: %s" % name) problems += 1 for name in sorted(was - now): print("no longer conditional, the note can go: %s" % name) problems += 1 sys.exit(1 if problems else 0) scopes = list(hc.SCOPES) if args.scope == "both" else [args.scope] picked = [e for e in entries if e["scope"] in scopes] if args.documented: picked = [e for e in picked if e["name"] in docs]