From 8080eef139b5dcf6201e4f27076a879d0df20096 Mon Sep 17 00:00:00 2001
From: Bryan Newbold <bnewbold@robocracy.org>
Date: Tue, 23 Nov 2021 18:58:37 -0800
Subject: file de-dupe: notes on prep and QA testing

---
 notes/cleanups/file_sha1_dedupe.md          | 64 +++++++++++++++++++++++++
 notes/cleanups/scripts/file_dupe_to_json.py | 72 +++++++++++++++++++++++++++++
 2 files changed, 136 insertions(+)
 create mode 100644 notes/cleanups/file_sha1_dedupe.md
 create mode 100755 notes/cleanups/scripts/file_dupe_to_json.py

(limited to 'notes')

diff --git a/notes/cleanups/file_sha1_dedupe.md b/notes/cleanups/file_sha1_dedupe.md
new file mode 100644
index 00000000..0829bc79
--- /dev/null
+++ b/notes/cleanups/file_sha1_dedupe.md
@@ -0,0 +1,64 @@
+
+
+## Prep
+
+Using `check_hashes.sh`:
+
+    zcat $HASH_FILE \
+        | awk '{print $3 "\t" $1}' \
+        | rg -v '^\t' \
+        | sort -S 4G \
+        | uniq -D -w 40 \
+        > sha1_ident.dupes.tsv
+
+    wc -l sha1_ident.dupes.tsv 
+    # 6,350
+
+    cut -f1 sha1_ident.dupes.tsv | uniq | wc -l
+    # 2,039
+
+Want to create JSON for each group, like:
+
+    entity_type: "file"
+    primary_id: str or None
+    duplicate_ids: [str]
+    evidence:
+        extid: str
+        extid_type: "sha1"
+
+Run transform script:
+
+    cat sha1_ident.dupes.tsv | ./file_dupe_to_json.py | pv -l > file_sha1_dupes.json
+    # 2.04k 0:00:00 [9.16k/s]
+
+
+## QA Testing
+
+    export FATCAT_AUTH_API_TOKEN=[...]
+
+    head -n25 /srv/fatcat/datasets/file_sha1_dupes.json \
+        | python -m fatcat_tools.mergers.files --editgroup-description-override "Automated merging of file entities with duplicate SHA-1 hashes" --dry-run merge-files -
+
+Hit some small bugs running in QA; test coverage isn't great, but I think hits
+the important parts.
+
+    head -n25 /srv/fatcat/datasets/file_sha1_dupes.json \
+        | python -m fatcat_tools.mergers.files --editgroup-description-override "Automated merging of file entities with duplicate SHA-1 hashes" --dry-run merge-files -
+    # Running in dry-run mode!
+    # Counter({'updated-entities': 60, 'lines': 25, 'merged': 25, 'skip': 0, 'updated-total': 0})
+
+Dry-run mode didn't actually work, and edits actually happened (!).
+
+Edits do look good.
+
+Try again, not dry-run, to ensure that case is handled:
+
+    head -n25 /srv/fatcat/datasets/file_sha1_dupes.json | python -m fatcat_tools.mergers.files --editgroup-description-override "Automated merging of file entities with duplicate SHA-1 hashes" merge-files -
+    # Counter({'lines': 25, 'skip': 25, 'skip-not-active-entity': 25, 'merged': 0, 'updated-total': 0})
+
+And then run 500 through for more testing:
+
+    head -n500 /srv/fatcat/datasets/file_sha1_dupes.json | python -m fatcat_tools.mergers.files --editgroup-description-override "Automated merging of file entities with duplicate SHA-1 hashes" merge-files -
+    # Counter({'updated-entities': 1341, 'lines': 500, 'merged': 474, 'skip': 26, 'skip-not-active-entity': 25, 'skip-entity-not-found': 1, 'updated-total': 0})
+
+The majority of merges seem to be cases where there are multiple articles in the same PDF.
diff --git a/notes/cleanups/scripts/file_dupe_to_json.py b/notes/cleanups/scripts/file_dupe_to_json.py
new file mode 100755
index 00000000..2064dc1c
--- /dev/null
+++ b/notes/cleanups/scripts/file_dupe_to_json.py
@@ -0,0 +1,72 @@
+#!/usr/bin/env python3
+
+"""
+This script can be used to transform duplicate file entity hash export rows
+into JSON objects which can be passed to the file entity merger.
+
+The input is expected to be a TSV with two columns: a hash value in the first
+column, and a fatcat file entity ident (in UUID format, not "fatcat ident"
+encoded) in the second column. The rows are assumed to be sorted by hash value
+(the first column), and duplicate values (same hash, differing UUID) are
+contiguous.
+
+File hashes aren't really "external identifiers" (ext_id), but we treat them as
+such here.
+
+Script is pretty simple, should be possible to copy and reuse for release,
+container, creator entity duplicates.
+"""
+
+import json, sys
+from typing import Optional
+import base64, uuid
+
+EXTID_TYPE = "sha1"
+
+def uuid2fcid(s: str) -> str:
+    """
+    Converts a uuid.UUID object to a fatcat identifier (base32 encoded string)
+    """     
+    raw = uuid.UUID(s).bytes
+    return base64.b32encode(raw)[:26].lower().decode("utf-8")
+
+def print_group(extid, dupe_ids):
+    if len(dupe_ids) < 2:
+        return
+    group = dict(
+        entity_type="file",
+        primary_id=None,
+        duplicate_ids=dupe_ids,
+        evidence=dict(
+            extid=extid,
+            extid_type=EXTID_TYPE,
+        ),
+    )
+    print(json.dumps(group, sort_keys=True))
+
+def run():
+    last_extid = None
+    dupe_ids = []
+    for l in sys.stdin:
+        l = l.strip()
+        if not l:
+            continue
+        (row_extid, row_uuid) = l.split("\t")[0:2]
+        if EXTID_TYPE == "sha1":
+            assert len(row_extid) == 40
+        else:
+            raise Exception(f"extid type not supported yet: {EXTID_TYPE}")
+        row_id = uuid2fcid(row_uuid)
+        if row_extid == last_extid:
+            dupe_ids.append(row_id)
+            continue
+        elif dupe_ids:
+            print_group(last_extid, dupe_ids)
+        last_extid = row_extid
+        dupe_ids = [row_id]
+    if last_extid and dupe_ids:
+        print_group(last_extid, dupe_ids)
+
+
+if __name__=="__main__":
+    run()
-- 
cgit v1.2.3