Emergent:Blu/Patches/recording-credits
From PickiPedia: A knowledge base of bluegrass, old time psychedelic jams, and other public domain music
Jump to navigationJump to search
Patch series for blu's branch blu/recording-credits in cryptograss/pickipedia: commits 01601ea and 5af0e4e, made with git format-patch origin/production..blu/recording-credits at block 26,286,602. Applies cleanly on production at 876f437 (checked with git am). sha256 of the file: 9b7c1d5932d6ecae883703815aa2e4032f3457dc86cef77fcb74b774ea86c9af. Fetched back from this page with action=raw and the text between the nowiki tags extracted, it hashes to b3d06f9a92b017fdc5efd92824c8fc7814489860dda0505a89523be8f33fc219: the only difference is the final blank line, which the wiki trims, and git am applies it cleanly on production at 876f437. For magent to git am and open the PR against production.
From 01601ea3c18e6aed2a4129d2b0d4c3ab0c69b819 Mon Sep 17 00:00:00 2001
From: blu <blu@cryptograss.live>
Date: Mon, 26 Oct 2026 15:32:18 +0000
Subject: [PATCH 1/2] Add recording_credits: who plays on a recording, from
MusicBrainz and Discogs
An Artificial Dumb lookup for the connectome. Given an artist and a title,
it finds every MusicBrainz recording, folds reissues to one release per
release group, and prints each source's performance credits marked by
scope: credited on this track, or on the whole release. Release-wide
credits are leads to check against liner notes, never answers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
tools/recording_credits.py | 271 ++++++++++++++++++++++++++++++++
tools/test_recording_credits.py | 152 ++++++++++++++++++
2 files changed, 423 insertions(+)
create mode 100644 tools/recording_credits.py
create mode 100644 tools/test_recording_credits.py
diff --git a/tools/recording_credits.py b/tools/recording_credits.py
new file mode 100644
index 0000000..468e3c5
--- /dev/null
+++ b/tools/recording_credits.py
@@ -0,0 +1,271 @@
+#!/usr/bin/env python3
+"""
+Who plays on a recording: an Artificial Dumb lookup.
+
+ python3 tools/recording_credits.py "Tony Rice" "House of the Rising Sun"
+
+Asks MusicBrainz for every recording of the title whose artist credit
+mentions the artist, then reads each release's credits from MusicBrainz and,
+where MusicBrainz links one, from Discogs. It prints what each source says,
+and nothing else: no merging of people across sources and no guessing.
+
+The scope of each credit matters as much as the credit. Both databases mostly
+attach players to the whole release ("tracks: all"), which says who is on the
+record, not who plays on this song. So every credit is marked:
+
+ track the source attaches it to this track
+ release the source attaches it to the whole release
+
+Only "track" credits answer the question. "release" credits are a lead to
+check against liner notes, and the output says so instead of
+promoting them.
+
+Reissues are folded together: one release per MusicBrainz release group,
+the earliest dated one.
+"""
+
+import json
+import re
+import sys
+import time
+from urllib.parse import urlencode
+from urllib.error import HTTPError
+from urllib.request import Request, urlopen
+
+USER_AGENT = "pickipedia-recording-credits/0.1 (https://pickipedia.xyz)"
+MB = "https://musicbrainz.org/ws/2"
+DISCOGS = "https://api.discogs.com"
+
+# MusicBrainz asks for at most one request a second; Discogs allows 25 a
+# minute without a key.
+MB_INTERVAL = 1.1
+DISCOGS_INTERVAL = 2.5
+
+# MusicBrainz relationship types that mean someone performed.
+MB_PERFORMANCE_TYPES = {"instrument", "vocal", "performer", "performing orchestra"}
+
+# Discogs roles that mean someone did something other than perform. Discogs
+# roles are free text, so this list is the dumb part: a role not on it is
+# shown as a performance. Extend it when a new one turns up.
+DISCOGS_OTHER_ROLES = (
+ "producer", "executive-producer", "co-producer", "mixed by", "mastered by",
+ "recorded by", "engineer", "liner notes", "design", "layout", "photography",
+ "art direction", "artwork", "written-by", "written by", "arranged by",
+ "composed by", "lyrics by", "music by", "lacquer cut", "coordinator",
+ "management", "booking", "edited by", "remastered by", "transferred by",
+ "notes", "illustration", "cover", "a&r", "copyright", "licensed",
+)
+
+_last_call = {}
+
+
+def _get_json(url, host, interval, attempts=4):
+ """GET a JSON document, pacing calls to each host.
+
+ Both services answer 503 (MusicBrainz) or 429 (Discogs) when we are going
+ too fast; those get retried with a growing wait. Anything else is raised.
+ """
+ req = Request(url, headers={"User-Agent": USER_AGENT, "Accept": "application/json"})
+ for attempt in range(attempts):
+ wait = _last_call.get(host, 0) + interval * (2 ** attempt) - time.monotonic()
+ if wait > 0:
+ time.sleep(wait)
+ try:
+ with urlopen(req, timeout=30) as resp:
+ return json.load(resp)
+ except HTTPError as e:
+ if e.code not in (429, 503) or attempt == attempts - 1:
+ raise
+ finally:
+ _last_call[host] = time.monotonic()
+
+
+def mb_get(path, **params):
+ params["fmt"] = "json"
+ return _get_json(f"{MB}/{path}?{urlencode(params)}", "mb", MB_INTERVAL)
+
+
+def discogs_get(path):
+ return _get_json(f"{DISCOGS}/{path}", "discogs", DISCOGS_INTERVAL)
+
+
+def normalize_title(title):
+ """Lowercase, straighten quotes, drop punctuation and a leading article."""
+ t = title.lower().replace("’", "'")
+ t = re.sub(r"[^\w\s']", " ", t)
+ t = re.sub(r"\s+", " ", t).strip()
+ return re.sub(r"^(the|a|an) ", "", t)
+
+
+def strip_discogs_suffix(name):
+ """Discogs disambiguates names as 'Craig Miller (3)'; the number isn't the name."""
+ return re.sub(r"\s\(\d+\)$", "", name)
+
+
+def credit_line(artist_credit):
+ return "".join(a["name"] + a.get("joinphrase", "") for a in artist_credit)
+
+
+def find_recordings(artist, title):
+ """Every MusicBrainz recording of title whose artist credit mentions artist."""
+ query = f'recording:"{title}" AND artist:"{artist}"'
+ data = mb_get("recording/", query=query, limit=100)
+ want = normalize_title(title)
+ return [
+ r for r in data.get("recordings", [])
+ if normalize_title(r["title"]) == want
+ and artist.lower() in credit_line(r["artist-credit"]).lower()
+ ]
+
+
+def pick_releases(recordings):
+ """One release per release group, the earliest dated; undated ones last."""
+ groups = {}
+ for rec in recordings:
+ for rel in rec.get("releases", []):
+ group = rel.get("release-group", {}).get("id", rel["id"])
+ key = (rel.get("date") or "9999", rel["id"])
+ if group not in groups or key < groups[group][0]:
+ groups[group] = (key, rec["id"], rel["id"])
+ return [(rec_id, rel_id) for _, rec_id, rel_id in sorted(groups.values())]
+
+
+def mb_credits(release, recording_id):
+ """Performance credits from a MusicBrainz release, with their scope."""
+ credits = []
+ for rel in release.get("relations", []):
+ if rel["type"] in MB_PERFORMANCE_TYPES and "artist" in rel:
+ credits.append(_mb_credit(rel, "release"))
+ track = None
+ for medium in release.get("media", []):
+ for t in medium.get("tracks", []):
+ if t["recording"]["id"] == recording_id:
+ track = t
+ for rel in t["recording"].get("relations", []):
+ if rel["type"] in MB_PERFORMANCE_TYPES and "artist" in rel:
+ credits.append(_mb_credit(rel, "track"))
+ return track, credits
+
+
+def _mb_credit(rel, scope):
+ role = ", ".join(rel.get("attributes") or []) or rel["type"]
+ return {"name": rel["artist"]["name"], "role": role, "scope": scope, "source": "MusicBrainz"}
+
+
+def discogs_id(release):
+ for rel in release.get("relations", []):
+ m = re.match(r"https?://(?:www\.)?discogs\.com/release/(\d+)", rel.get("url", {}).get("resource", ""))
+ if m:
+ return m.group(1)
+ return None
+
+
+def track_in_spec(position, spec):
+ """Whether a Discogs 'tracks' spec such as '1, 3 to 5' or 'A2' covers position."""
+ for part in (p.strip() for p in spec.split(",")):
+ if " to " in part:
+ lo, hi = (x.strip() for x in part.split(" to ", 1))
+ if lo.isdigit() and hi.isdigit() and position.isdigit():
+ if int(lo) <= int(position) <= int(hi):
+ return True
+ elif part == position:
+ return True
+ elif part == position:
+ return True
+ return False
+
+
+def is_performance(role):
+ r = role.lower()
+ return not any(r.startswith(other) for other in DISCOGS_OTHER_ROLES)
+
+
+def discogs_credits(release, title):
+ """Performance credits from a Discogs release for the track titled title."""
+ want = normalize_title(title)
+ track = next((t for t in release.get("tracklist", []) if normalize_title(t["title"]) == want), None)
+ credits = []
+ position = track["position"] if track else None
+ for e in release.get("extraartists", []):
+ if not is_performance(e["role"]):
+ continue
+ spec = (e.get("tracks") or "").strip()
+ if not spec:
+ scope = "release"
+ elif position and track_in_spec(position, spec):
+ scope = "track"
+ else:
+ continue # credited on other tracks only
+ credits.append({"name": strip_discogs_suffix(e["name"]), "role": e["role"],
+ "scope": scope, "source": "Discogs"})
+ for e in (track or {}).get("extraartists", []):
+ if is_performance(e["role"]):
+ credits.append({"name": strip_discogs_suffix(e["name"]), "role": e["role"],
+ "scope": "track", "source": "Discogs"})
+ return track, credits
+
+
+def lookup(artist, title):
+ """Everything the sources say about who plays on each version."""
+ versions = []
+ for recording_id, release_id in pick_releases(find_recordings(artist, title)):
+ release = mb_get(f"release/{release_id}",
+ inc="artist-credits+labels+recordings+artist-rels+recording-level-rels+url-rels")
+ track, credits = mb_credits(release, recording_id)
+ version = {
+ "artist": credit_line(release.get("artist-credit", [])),
+ "release": release["title"],
+ "date": release.get("date", ""),
+ "labels": [f'{li["label"]["name"]} {li.get("catalog-number") or ""}'.strip()
+ for li in release.get("label-info", []) if li.get("label")],
+ "track": track["position"] if track else None,
+ "sources": [f"https://musicbrainz.org/release/{release_id}"],
+ "credits": credits,
+ }
+ did = discogs_id(release)
+ if did:
+ dtrack, dcredits = discogs_credits(discogs_get(f"releases/{did}"), title)
+ version["sources"].append(f"https://www.discogs.com/release/{did}")
+ version["credits"] += dcredits
+ if dtrack is None:
+ version["note"] = "Discogs tracklist has no track with this title"
+ versions.append(version)
+ return versions
+
+
+def render(artist, title, versions):
+ lines = [f'Who plays on {artist}: "{title}"', ""]
+ if not versions:
+ lines.append("MusicBrainz has no recording with this title and artist.")
+ for v in versions:
+ lines.append(f'{v["artist"]}: {v["release"]} ({v["date"] or "undated"}), '
+ f'{"; ".join(v["labels"]) or "no label"}, track {v["track"]}')
+ for scope, heading in (("track", "Credited on this track"),
+ ("release", "Credited on the whole release (check liner notes)")):
+ rows = [c for c in v["credits"] if c["scope"] == scope]
+ lines.append(f" {heading}:")
+ if not rows:
+ lines.append(" (none)")
+ for c in rows:
+ lines.append(f' {c["name"]}: {c["role"]} [{c["source"]}]')
+ if v.get("note"):
+ lines.append(f' Note: {v["note"]}')
+ lines.append(" Sources: " + " ".join(v["sources"]))
+ lines.append("")
+ return "\n".join(lines)
+
+
+def main(argv):
+ args = [a for a in argv[1:] if a != "--json"]
+ if len(args) != 2:
+ print(__doc__.strip().splitlines()[0])
+ print('usage: recording_credits.py [--json] "ARTIST" "TITLE"')
+ return 2
+ artist, title = args
+ versions = lookup(artist, title)
+ print(json.dumps(versions, indent=2) if "--json" in argv else render(artist, title, versions))
+ return 0
+
+
+if __name__ == "__main__":
+ sys.exit(main(sys.argv))
diff --git a/tools/test_recording_credits.py b/tools/test_recording_credits.py
new file mode 100644
index 0000000..e96055b
--- /dev/null
+++ b/tools/test_recording_credits.py
@@ -0,0 +1,152 @@
+#!/usr/bin/env python3
+"""
+Tests for the recording-credits lookup.
+
+The thing these guard is scope: a credit the source attaches to the whole
+release must never be reported as a credit on the track. Everything here runs
+on canned documents shaped like the real MusicBrainz and Discogs responses,
+so no test touches the network.
+"""
+
+import pytest
+
+import recording_credits as rc
+
+RECORDING = "rec-1"
+
+
+def mb_release(release_relations=(), track_relations=(), url=None):
+ rels = [{"type": t, "attributes": attrs, "artist": {"name": n}} for t, attrs, n in release_relations]
+ if url:
+ rels.append({"type": "discogs", "url": {"resource": url}})
+ return {
+ "title": "Unit of Measure", "date": "2000", "relations": rels,
+ "artist-credit": [{"name": "The Tony Rice Unit", "joinphrase": ""}],
+ "label-info": [{"label": {"name": "Rounder"}, "catalog-number": "11661-0405-2"}],
+ "media": [{"tracks": [
+ {"position": 1, "recording": {"id": "other", "relations": [
+ {"type": "instrument", "attributes": ["banjo"], "artist": {"name": "Someone Else"}}]}},
+ {"position": 2, "recording": {"id": RECORDING, "relations": [
+ {"type": t, "attributes": attrs, "artist": {"name": n}} for t, attrs, n in track_relations]}},
+ ]}],
+ }
+
+
+# ---------------------------------------------------------------------------
+# MusicBrainz scope
+# ---------------------------------------------------------------------------
+
+def test_release_relations_are_release_scope():
+ track, credits = rc.mb_credits(mb_release([("instrument", ["mandolin"], "Jimmy Gaudreau")]), RECORDING)
+ assert track["position"] == 2
+ assert credits == [{"name": "Jimmy Gaudreau", "role": "mandolin", "scope": "release", "source": "MusicBrainz"}]
+
+
+def test_track_relations_are_track_scope_and_other_tracks_ignored():
+ _, credits = rc.mb_credits(mb_release(track_relations=[("vocal", ["lead vocals"], "Jerry Garcia")]), RECORDING)
+ assert credits == [{"name": "Jerry Garcia", "role": "lead vocals", "scope": "track", "source": "MusicBrainz"}]
+
+
+def test_non_performance_relations_are_dropped():
+ _, credits = rc.mb_credits(mb_release([("producer", [], "Bill Wolf"), ("mix", [], "Bill Wolf")]), RECORDING)
+ assert credits == []
+
+
+def test_discogs_id_from_url_relation():
+ assert rc.discogs_id(mb_release(url="https://www.discogs.com/release/4090806")) == "4090806"
+ assert rc.discogs_id(mb_release()) is None
+
+
+# ---------------------------------------------------------------------------
+# Discogs scope
+# ---------------------------------------------------------------------------
+
+def discogs_release(extraartists, track_extras=()):
+ return {
+ "extraartists": extraartists,
+ "tracklist": [
+ {"position": "1", "title": "Old Train", "extraartists": []},
+ {"position": "2", "title": "House Of The Rising Sun", "extraartists": list(track_extras)},
+ {"position": "3", "title": "Something Else", "extraartists": []},
+ ],
+ }
+
+
+def test_discogs_empty_tracks_is_release_scope():
+ _, credits = rc.discogs_credits(discogs_release([{"name": "Tony Rice", "role": "Guitar", "tracks": ""}]),
+ "House of the Rising Sun")
+ assert [(c["name"], c["scope"]) for c in credits] == [("Tony Rice", "release")]
+
+
+@pytest.mark.parametrize("spec,included", [
+ ("2", True), ("1, 2", True), ("1 to 3", True), ("3", False), ("1, 3", False), ("3 to 9", False),
+])
+def test_discogs_tracks_spec(spec, included):
+ _, credits = rc.discogs_credits(
+ discogs_release([{"name": "Guest Player", "role": "Dobro", "tracks": spec}]), "House of the Rising Sun")
+ assert [c["scope"] for c in credits] == (["track"] if included else [])
+
+
+def test_discogs_track_level_credits_and_suffix_stripped():
+ _, credits = rc.discogs_credits(
+ discogs_release([], [{"name": "Craig Miller (3)", "role": "Harmonica"},
+ {"name": "Jerry Garcia", "role": "Arranged By"}]),
+ "House of the Rising Sun")
+ assert credits == [{"name": "Craig Miller", "role": "Harmonica", "scope": "track", "source": "Discogs"}]
+
+
+def test_discogs_production_roles_are_not_performances():
+ for role in ("Producer", "Mixed By, Mastered By", "Liner Notes [Song Notes]", "Recorded By", "Photography By"):
+ assert not rc.is_performance(role), role
+ for role in ("Guitar", "Acoustic Bass", "Guitar, Vocals", "Fiddle"):
+ assert rc.is_performance(role), role
+
+
+def test_discogs_missing_track_keeps_release_credits_only():
+ track, credits = rc.discogs_credits(
+ discogs_release([{"name": "Tony Rice", "role": "Guitar", "tracks": ""},
+ {"name": "Guest Player", "role": "Dobro", "tracks": "2"}]),
+ "A Song Not On It")
+ assert track is None
+ assert [(c["name"], c["scope"]) for c in credits] == [("Tony Rice", "release")]
+
+
+# ---------------------------------------------------------------------------
+# Finding versions
+# ---------------------------------------------------------------------------
+
+def test_normalize_title():
+ assert rc.normalize_title("The House Of The Rising Sun") == rc.normalize_title("House of the Rising Sun")
+ assert rc.normalize_title("Can’t Outrun the Blues") == "can't outrun the blues"
+
+
+def test_find_recordings_filters_title_and_artist(monkeypatch):
+ monkeypatch.setattr(rc, "mb_get", lambda path, **params: {"recordings": [
+ {"id": "a", "title": "House of the Rising Sun", "artist-credit": [{"name": "The Tony Rice Unit"}]},
+ {"id": "b", "title": "House of the Rising Sun Blues", "artist-credit": [{"name": "Tony Rice"}]},
+ {"id": "c", "title": "House of the Rising Sun", "artist-credit": [{"name": "Larry Rice"}]},
+ ]})
+ assert [r["id"] for r in rc.find_recordings("Tony Rice", "House of the Rising Sun")] == ["a"]
+
+
+def test_pick_releases_one_per_group_earliest_first():
+ recordings = [
+ {"id": "r1", "releases": [
+ {"id": "reissue", "date": "2011-10-18", "release-group": {"id": "g1"}},
+ {"id": "original", "date": "2000", "release-group": {"id": "g1"}}]},
+ {"id": "r2", "releases": [
+ {"id": "undated", "release-group": {"id": "g2"}},
+ {"id": "pizza", "date": "2000-04-25", "release-group": {"id": "g2"}}]},
+ ]
+ assert rc.pick_releases(recordings) == [("r1", "original"), ("r2", "pizza")]
+
+
+def test_render_separates_track_from_release():
+ out = rc.render("Tony Rice", "House of the Rising Sun", [{
+ "artist": "The Tony Rice Unit", "release": "Unit of Measure", "date": "2000",
+ "labels": ["Rounder 11661-0405-2"], "track": 2, "sources": ["https://musicbrainz.org/release/x"],
+ "credits": [{"name": "Tony Rice", "role": "guitar", "scope": "release", "source": "MusicBrainz"}],
+ }])
+ track_part, release_part = out.split("Credited on the whole release")
+ assert "(none)" in track_part and "Tony Rice" not in track_part.split("track 2", 1)[1]
+ assert "Tony Rice: guitar [MusicBrainz]" in release_part
--
2.53.0
From 5af0e4e289d0de86502429c6254dbde420866770 Mon Sep 17 00:00:00 2001
From: blu <blu@cryptograss.live>
Date: Tue, 27 Oct 2026 17:04:19 +0000
Subject: [PATCH 2/2] recording_credits: judge each part of a composite Discogs
role
Traversal outward from Unit of Measure turned up two leaks in the
non-performance filter, which judged the whole role string:
- 'Guitar, Producer' starts with 'guitar', so producers were listed as
players (Tony Rice on Got Me a Martin Guitar; Scott Vestal on Bluegrass '97).
- 'Producer, Engineer, Recorded By, Guitar' starts with 'producer', so two
players were hidden entirely (Wyatt Rice and Larry Rice on Larry and Wyatt
Rice).
Roles are now split on commas and slashes and each part is judged alone; the
printed role keeps only the performance parts. The role list is normalized on
both sides, so 'Executive Producer' and 'Executive-Producer' are one entry,
and the roles the traversal exposed are added: songwriter, sleeve notes,
created by, film director, music director, project manager.
Releases with the same title and date in different release groups fold into
one. Differently-titled releases of one show do not: nothing in the data says
they are one tape.
---
tools/recording_credits.py | 78 ++++++++++++++++++++++++++-------
tools/test_recording_credits.py | 53 ++++++++++++++++++++++
2 files changed, 115 insertions(+), 16 deletions(-)
diff --git a/tools/recording_credits.py b/tools/recording_credits.py
index 468e3c5..bc36e2b 100644
--- a/tools/recording_credits.py
+++ b/tools/recording_credits.py
@@ -44,17 +44,35 @@ DISCOGS_INTERVAL = 2.5
# MusicBrainz relationship types that mean someone performed.
MB_PERFORMANCE_TYPES = {"instrument", "vocal", "performer", "performing orchestra"}
+def _norm_role(text):
+ """A Discogs role reduced to bare words: 'Lead Guitar [Electric]' -> 'lead guitar'.
+
+ Discogs writes the same job several ways -- 'Executive Producer' and
+ 'Executive-Producer', 'Written By' and 'Written-By' -- so both the list
+ below and each role being tested go through here before they are compared.
+ The bracketed part is a qualifier ('Liner Notes [Song Notes]'), never the job.
+ """
+ text = re.sub(r"\[[^\]]*\]|\([^)]*\)", " ", text.lower())
+ return re.sub(r"\s+", " ", re.sub(r"[^\w\s]", " ", text)).strip()
+
+
# Discogs roles that mean someone did something other than perform. Discogs
# roles are free text, so this list is the dumb part: a role not on it is
-# shown as a performance. Extend it when a new one turns up.
-DISCOGS_OTHER_ROLES = (
- "producer", "executive-producer", "co-producer", "mixed by", "mastered by",
- "recorded by", "engineer", "liner notes", "design", "layout", "photography",
- "art direction", "artwork", "written-by", "written by", "arranged by",
- "composed by", "lyrics by", "music by", "lacquer cut", "coordinator",
- "management", "booking", "edited by", "remastered by", "transferred by",
- "notes", "illustration", "cover", "a&r", "copyright", "licensed",
-)
+# shown as a performance. Extend it when a new one turns up; the entries are
+# normalized the same way a role is, so write them however reads best.
+DISCOGS_OTHER_ROLES = tuple(_norm_role(r) for r in (
+ "producer", "executive producer", "co producer", "associate producer",
+ "mixed by", "mixing", "mastered by", "mastering", "recorded by",
+ "engineer", "liner notes", "sleeve notes", "song notes", "notes",
+ "design", "layout", "graphic design", "art direction", "artwork",
+ "illustration", "cover", "photography", "photography by", "photographer",
+ "written by", "songwriter", "composed by", "lyrics by", "music by",
+ "arranged by", "lacquer cut", "coordinator", "project coordination",
+ "project manager", "production manager", "management", "booking",
+ "edited by", "editor", "remastered by", "transferred by", "compiled by",
+ "a&r", "copyright", "licensed", "manufactured by", "distributed by",
+ "created by", "directed by", "film director", "music director", "legal",
+))
_last_call = {}
@@ -119,15 +137,26 @@ def find_recordings(artist, title):
def pick_releases(recordings):
- """One release per release group, the earliest dated; undated ones last."""
+ """One release per release group, the earliest dated; undated ones last.
+
+ Releases outside a group that share a title and date are folded too: the
+ same show turns up as several release groups, sometimes one per taper.
+ """
groups = {}
for rec in recordings:
for rel in rec.get("releases", []):
group = rel.get("release-group", {}).get("id", rel["id"])
key = (rel.get("date") or "9999", rel["id"])
if group not in groups or key < groups[group][0]:
- groups[group] = (key, rec["id"], rel["id"])
- return [(rec_id, rel_id) for _, rec_id, rel_id in sorted(groups.values())]
+ groups[group] = (key, rec["id"], rel["id"], rel)
+ seen, picked = set(), []
+ for key, rec_id, rel_id, rel in sorted(groups.values()):
+ shown = (normalize_title(rel.get("title", "")), rel.get("date") or "")
+ if shown in seen:
+ continue
+ seen.add(shown)
+ picked.append((rec_id, rel_id))
+ return picked
def mb_credits(release, recording_id):
@@ -175,9 +204,24 @@ def track_in_spec(position, spec):
return False
+def performance_roles(role):
+ """The parts of a Discogs role that mean performing.
+
+ A role is often several jobs at once -- 'Banjo, Producer, Mixed By' -- so
+ each part is judged on its own. Judging the whole string would let the
+ first word hide the rest: 'Guitar, Producer' starts with 'guitar', and
+ reporting it as a performance would credit a producer as a player.
+ """
+ parts = []
+ for part in re.split(r"[,/]", role):
+ norm = _norm_role(part)
+ if norm and not any(norm.startswith(other) for other in DISCOGS_OTHER_ROLES):
+ parts.append(part.strip())
+ return parts
+
+
def is_performance(role):
- r = role.lower()
- return not any(r.startswith(other) for other in DISCOGS_OTHER_ROLES)
+ return bool(performance_roles(role))
def discogs_credits(release, title):
@@ -196,11 +240,13 @@ def discogs_credits(release, title):
scope = "track"
else:
continue # credited on other tracks only
- credits.append({"name": strip_discogs_suffix(e["name"]), "role": e["role"],
+ credits.append({"name": strip_discogs_suffix(e["name"]),
+ "role": ", ".join(performance_roles(e["role"])),
"scope": scope, "source": "Discogs"})
for e in (track or {}).get("extraartists", []):
if is_performance(e["role"]):
- credits.append({"name": strip_discogs_suffix(e["name"]), "role": e["role"],
+ credits.append({"name": strip_discogs_suffix(e["name"]),
+ "role": ", ".join(performance_roles(e["role"])),
"scope": "track", "source": "Discogs"})
return track, credits
diff --git a/tools/test_recording_credits.py b/tools/test_recording_credits.py
index e96055b..05f2f34 100644
--- a/tools/test_recording_credits.py
+++ b/tools/test_recording_credits.py
@@ -150,3 +150,56 @@ def test_render_separates_track_from_release():
track_part, release_part = out.split("Credited on the whole release")
assert "(none)" in track_part and "Tony Rice" not in track_part.split("track 2", 1)[1]
assert "Tony Rice: guitar [MusicBrainz]" in release_part
+
+
+# ---------------------------------------------------------------------------
+# Composite roles
+#
+# A Discogs role can name several jobs at once. The first version of the
+# filter judged the whole string, so 'Guitar, Producer' counted as a
+# performance and put a producer in the player list.
+# ---------------------------------------------------------------------------
+
+@pytest.mark.parametrize("role,performing", [
+ ("Guitar, Producer", ["Guitar"]),
+ ("Banjo, Producer, Mixed By", ["Banjo"]),
+ ("Vocals, Guitar", ["Vocals", "Guitar"]),
+ ("Producer, Guitar", ["Guitar"]),
+ ("Executive Producer", []),
+ ("Executive-Producer", []),
+ ("Songwriter", []),
+ ("Sleeve Notes", []),
+ ("Created By, Executive-Producer", []),
+ ("Film Director [Directed By]", []),
+ ("Project Manager [Project Coordination]", []),
+ ("Liner Notes [Song Notes]", []),
+ ("Written-By", []),
+])
+def test_composite_roles(role, performing):
+ assert rc.performance_roles(role) == performing
+
+
+def test_producer_hidden_behind_a_performance_is_not_reported():
+ _, credits = rc.discogs_credits(
+ discogs_release([{"name": "Tony Rice", "role": "Guitar, Producer", "tracks": ""}],
+ [{"name": "Scott Vestal", "role": "Banjo, Producer, Mixed By"}]),
+ "House of the Rising Sun")
+ assert [(c["name"], c["role"], c["scope"]) for c in credits] == [
+ ("Tony Rice", "Guitar", "release"), ("Scott Vestal", "Banjo", "track")]
+
+
+def test_literal_duplicates_across_groups_are_folded():
+ """Same title and date in two groups: one entry.
+
+ Differently-titled releases of one show are NOT folded: 'Warfield
+ Theater, 12-08-91 (disc 1)' and '1991-12-08 Warfield Theater, SF, CA'
+ stay two versions, because nothing in the data says they are one tape.
+ """
+ recordings = [{"id": "r1", "releases": [
+ {"id": "a", "title": "Shady Grove", "date": "1996-10-29", "release-group": {"id": "g1"}},
+ {"id": "b", "title": "Shady Grove", "date": "1996-10-29", "release-group": {"id": "g2"}},
+ {"id": "c", "title": "Warfield Theater, 12-08-91 (disc 1)", "date": "1991-12-08",
+ "release-group": {"id": "g3"}},
+ {"id": "d", "title": "1991-12-08 Warfield Theater, SF, CA, USA", "date": "1991-12-08",
+ "release-group": {"id": "g4"}}]}]
+ assert rc.pick_releases(recordings) == [("r1", "c"), ("r1", "d"), ("r1", "a")]
--
2.53.0
