ai4c-scribe/metadiff-obo-config.yaml at main · ai4curation/ai4c-scribe · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
# Metadiff Configuration for OBO (Open Biomedical Ontology) Files
# ================================================================
#
# This configuration is optimized for comparing diffs in OBO format files
# (Gene Ontology, Human Phenotype Ontology, Mondo, etc.)
#
# Use case: Evaluate agents that fix ontology issues
# Example: Agent adds new term with different ID than human, should still match
#
# Key behaviors:
# - Masks arbitrary CURIE IDs (GO:0000001 → GO:NNNNNNN)
# - Ignores metadata fields that vary by author/time (created_by, creation_date)
# - Preserves semantic content (name, definition, relationships)
#
# Usage:
#   # CLI
#   ai4c-scribe metadiff compare human.obo agent.obo -c obo
#
#   # Python API
#   from ai4c_scribe.metadiff import compare_diff_files, get_config
#   result = compare_diff_files(Path("human.diff"), Path("agent.diff"), config=get_config("obo"))

name: "obo"
description: |
  OBO (Open Biomedical Ontology) file comparison.
  Masks arbitrary term IDs and ignores author/timestamp metadata.

normalizer:
  # Mask CURIE-format IDs (GO:0000001 -> GO:NNNNNNN)
  # This is crucial: new term IDs are semantically arbitrary.
  # We care about the term content (name, definition, relationships), not the ID.
  mask_ids: true

  # Strip leading/trailing whitespace from each line
  strip_whitespace: true

  # Not case-sensitive for OBO (IDs are case-sensitive, but content isn't)
  case_insensitive: false

  # Regex patterns to skip entirely (not compared at all)
  ignore_patterns: []

  # Exact substring matches to skip (if line contains any of these, skip it)
  ignore_keys:
    # Metadata that varies by author - not semantic
    - "created_by"
    - "dc-contributor"

    # Metadata that varies by time - not semantic
    - "creation_date"
    - "dcterms-date"
    - "property_value: dcterms-date"
    - "relationship: dc-contributor"

  # Custom normalizers to apply (in order)
  custom_normalizers:
    - mask_timestamps      # Normalize date formats
    - mask_version_numbers # Normalize version number changes

# Generate visual diff using icdiff (colored side-by-side comparison)
generate_visual: true

# Arguments to pass to icdiff command
icdiff_args: ["-E", "@@"]


# ================================================================
# EXAMPLES
# ================================================================
#
# Example 1: Identical terms with different IDs
# --------------------------------------------------
# Human PR:
#   +[Term]
#   +id: GO:0000001
#   +name: biological process
#   +def: "Process related to biology"
#
# Agent PR (different ID, same content):
#   +[Term]
#   +id: GO:0000999
#   +name: biological process
#   +def: "Process related to biology"
#
# Result with OBO config: similarity=1.0 (IDs are masked to GO:NNNNNNN)
#
#
# Example 2: Same term with different metadata
# --------------------------------------------------
# Human PR:
#   +[Term]
#   +id: GO:0000001
#   +name: protein binding
#   +created_by: alice
#   +creation_date: 2024-01-15
#
# Agent PR (same content, different author/date):
#   +[Term]
#   +id: GO:0000001
#   +name: protein binding
#   +created_by: bob
#   +creation_date: 2024-01-20
#
# Result with OBO config: similarity=1.0 (metadata ignored)
#
#
# Example 3: Different definitions (should be caught)
# --------------------------------------------------
# Human PR:
#   +name: kinase activity
#   +def: "Activity that transfers phosphate groups"
#
# Agent PR:
#   +name: kinase activity
#   +def: "Activity that transfers molecules"
#
# Result with OBO config: similarity<1.0 (definitions differ - SHOULD be caught)
#
# ================================================================

# For comparison with CODE diffs, use the 'python' config instead:
#   ai4c-scribe metadiff compare code.diff agent.diff -c python
#
# For strict byte-for-byte comparison, use:
#   ai4c-scribe metadiff compare file1.diff file2.diff -c strict