Merge branch 'martin-matcher-class' into 'master'

turn "match_release_fuzzy" into a class

See merge request webgroup/fuzzycat!10

16 files changed, 323 insertions, 12 deletions

diff --git a/tests/files/README.md b/tests/files/README.md
new file mode 100644
index 0000000..ef674d6
--- /dev/null
+++ b/tests/files/README.md
@@ -0,0 +1,5 @@
+# Matcher Test Files
+
+The goal here is to have a mostly language-independent test cases for matching.
+
+Each subdirectory corresponds to a test function and contains examples for it.
diff --git a/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/0.yaml b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/0.yaml
new file mode 100644
index 0000000..2df8d9a
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/0.yaml
@@ -0,0 +1,13 @@
+title: titles are case insensitive
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Michael Adams"
+      }
+    ],
+    "title": "digital libraries",
+    "ext_ids": {}
+  }
+expected: 2
diff --git a/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/1.yaml b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/1.yaml
new file mode 100644
index 0000000..1070408
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/1.yaml
@@ -0,0 +1,13 @@
+title: another vanilla query
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Poul-Henning Kamp"
+      }
+    ],
+    "title": "The hyperdimensional tar pit",
+    "ext_ids": {}
+  }
+expected: 2
diff --git a/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/2.yaml b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/2.yaml
new file mode 100644
index 0000000..882e746
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/2.yaml
@@ -0,0 +1,16 @@
+title: order of contribs does not matter
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Maurice Florence"
+      },
+      {
+        "raw_name": "Tuomo Tiisala"
+      }
+    ],
+    "title": "Foucault",
+    "ext_ids": {}
+  }
+expected: 1
diff --git a/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/3.yaml b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/3.yaml
new file mode 100644
index 0000000..0a2ad12
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/3.yaml
@@ -0,0 +1,16 @@
+title: order of contribs does not matter
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Tuomo Tiisala"
+      },
+      {
+        "raw_name": "Maurice Florence"
+      }
+    ],
+    "title": "Foucault",
+    "ext_ids": {}
+  }
+expected: 1
diff --git a/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/4.yaml b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/4.yaml
new file mode 100644
index 0000000..36ea0fe
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_exact_contrib/4.yaml
@@ -0,0 +1,16 @@
+title: short version of name should not work
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Tuomo Tiisala"
+      },
+      {
+        "raw_name": "M. Florence"
+      }
+    ],
+    "title": "Foucault",
+    "ext_ids": {}
+  }
+expected: 0
diff --git a/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/0.yaml b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/0.yaml
new file mode 100644
index 0000000..07230e8
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/0.yaml
@@ -0,0 +1,14 @@
+title: titles are case insensitive
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Michael Adams"
+      }
+    ],
+    "title": "digital libraries",
+    "ext_ids": {}
+  }
+jaccard_index_threshold: 1.0
+expected: 2
diff --git a/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/1.yaml b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/1.yaml
new file mode 100644
index 0000000..62e9586
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/1.yaml
@@ -0,0 +1,14 @@
+title: another vanilla query
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Poul-Henning Kamp"
+      }
+    ],
+    "title": "The hyperdimensional tar pit",
+    "ext_ids": {}
+  }
+jaccard_index_threshold: 1.0
+expected: 2
diff --git a/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/2.yaml b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/2.yaml
new file mode 100644
index 0000000..b89e825
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/2.yaml
@@ -0,0 +1,17 @@
+title: order of contribs does not matter
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Maurice Florence"
+      },
+      {
+        "raw_name": "Tuomo Tiisala"
+      }
+    ],
+    "title": "Foucault",
+    "ext_ids": {}
+  }
+jaccard_index_threshold: 1.0
+expected: 1
diff --git a/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/3.yaml b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/3.yaml
new file mode 100644
index 0000000..3de7262
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/3.yaml
@@ -0,0 +1,17 @@
+title: order of contribs does not matter
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Tuomo Tiisala"
+      },
+      {
+        "raw_name": "Maurice Florence"
+      }
+    ],
+    "title": "Foucault",
+    "ext_ids": {}
+  }
+jaccard_index_threshold: 1.0
+expected: 1
diff --git a/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/4.yaml b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/4.yaml
new file mode 100644
index 0000000..39fb065
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/4.yaml
@@ -0,0 +1,17 @@
+title: short version of name should not work
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Tuomo Tiisala"
+      },
+      {
+        "raw_name": "M. Florence"
+      }
+    ],
+    "title": "Foucault",
+    "ext_ids": {}
+  }
+jaccard_index_threshold: 1.0
+expected: 0
diff --git a/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/5.yaml b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/5.yaml
new file mode 100644
index 0000000..fff19fa
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/5.yaml
@@ -0,0 +1,17 @@
+title: here, Iz Beltagy is missing from author, but still retrieved
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Arman Cohan"
+      },
+      {
+        "raw_name": "Kyle Lo"
+      }
+    ],
+    "title": "SciBERT: A Pretrained Language Model for Scientific Text",
+    "ext_ids": {}
+  }
+jaccard_index_threshold: 0.5
+expected: 3
diff --git a/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/6.yaml b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/6.yaml
new file mode 100644
index 0000000..d4e0025
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/6.yaml
@@ -0,0 +1,14 @@
+title: here, 2/3 authors are missing, we fail with jaccard index 0.5
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Arman Cohan"
+      }
+    ],
+    "title": "SciBERT: A Pretrained Language Model for Scientific Text",
+    "ext_ids": {}
+  }
+jaccard_index_threshold: 0.5
+expected: 0
diff --git a/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/7.yaml b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/7.yaml
new file mode 100644
index 0000000..23d5a8d
--- /dev/null
+++ b/tests/files/fuzzy_release_match_release_exact_title_partial_contrib/7.yaml
@@ -0,0 +1,17 @@
+title: match, despite trailing whitespace
+date: 2021-11-08
+input: >
+  {
+    "contribs": [
+      {
+        "raw_name": "Arman Cohan"
+      },
+      {
+        "raw_name": "Kyle Lo"
+      }
+    ],
+    "title": "SciBERT: A Pretrained Language Model for Scientific Text    ",
+    "ext_ids": {}
+  }
+jaccard_index_threshold: 0.5
+expected: 3
diff --git a/tests/test_grobid_unstructured.py b/tests/test_grobid_unstructured.py
index cf71f91..f36f9a4 100644
--- a/tests/test_grobid_unstructured.py
+++ b/tests/test_grobid_unstructured.py
@@ -18,11 +18,7 @@ def test_grobid_ref_to_release():
                 given_name='ahab',
                 surname='sailor',
             ),
-            GrobidAuthor(
-                full_name='mary jane',
-                given_name='mary',
-                surname='jane'
-            ),
+            GrobidAuthor(full_name='mary jane', given_name='mary', surname='jane'),
         ],
     )
     r = grobid_ref_to_release(d)
diff --git a/tests/test_matching.py b/tests/test_matching.py
index ad971a5..ca94c2a 100644
--- a/tests/test_matching.py
+++ b/tests/test_matching.py
@@ -1,13 +1,14 @@
+import collections
 import logging
 import warnings
 
 import elasticsearch
 import pytest
 import requests
-from fatcat_openapi_client import ReleaseEntity
+from fatcat_openapi_client import ReleaseEntity, ReleaseContrib
 
-from fuzzycat.entities import entity_from_dict
-from fuzzycat.matching import anything_to_entity, match_release_fuzzy
+from fuzzycat.entities import entity_from_dict, entity_from_json
+from fuzzycat.matching import anything_to_entity, match_release_fuzzy, FuzzyReleaseMatcher
 
 warnings.filterwarnings(
     "ignore")  # InsecureRequestWarning: Unverified HTTPS request is being made to host ...
@@ -18,6 +19,9 @@ from fatcat_openapi_client import ReleaseEntity
 import pytest
 import elasticsearch
 import logging
+import yaml
+import glob
+import json
 
 logger = logging.getLogger('test_matching')
 logger.setLevel(logging.DEBUG)
@@ -40,19 +44,35 @@ def is_reachable(url, timeout=3):
         return False
 
 
+def yaml_to_cases(klass,
+                  files="tests/files/fuzzy_release_match_release_exact_title_exact_contrib/*.yaml"):
+    """
+    Turn yaml files into a collection of named tuple test cases. The glob is
+    relative to the project root (i.e. where you usually run `pytest` from).
+    """
+    cases = []
+    for path in glob.glob(files):
+        with open(path) as f:
+            doc = yaml.load(f, Loader=yaml.Loader)
+            cases.append(klass(**doc))
+    return cases
+
+
 @pytest.fixture
 def es_client():
     return elasticsearch.Elasticsearch([FATCAT_SEARCH_URL])
 
 
-@pytest.mark.skipif(
-    is_not_reachable(FATCAT_SEARCH_URL),
-    reason="{} not reachable, use e.g. FUZZYCAT_FATCAT_SEARCH_URL=localhost:9200 to override".
-    format(FATCAT_SEARCH_URL))
+# @pytest.mark.skipif(
+#     is_not_reachable(FATCAT_SEARCH_URL),
+#     reason="{} not reachable, use e.g. FUZZYCAT_FATCAT_SEARCH_URL=localhost:9200 to override".
+#     format(FATCAT_SEARCH_URL))
 def test_match_release_fuzzy(es_client, caplog):
     """
     This test is tied to the current index contents, so if that changes, this
     test may fail as well.
+
+    Note: Deprecated. We want to get rid of this.
     """
     cases = (
         ("wtv64ahbdzgwnan7rllwr3nurm", 1),
@@ -106,3 +126,92 @@ def test_match_release_fuzzy(es_client, caplog):
             logging.info("[{}] given title '{}', found {}, {}".format(i, entity.title, len(result),
                                                                       [v.title for v in result]))
         assert len(result) == count, doc
+
+
+def test_matcher_match_release(es_client, caplog):
+    cases = (
+        ("wtv64ahbdzgwnan7rllwr3nurm", 1),
+        ("eqcgtpav3na5jh56o5vjsvb4ei", 1),
+    )
+    matcher = FuzzyReleaseMatcher(es=es_client, size=5)
+    for i, (ident, count) in enumerate(cases):
+        entity = anything_to_entity(ident, ReleaseEntity)
+        result = matcher.match(entity)
+        logger.info("[{}] given {}, found {}".format(i, entity.title, len(result)))
+        assert len(result) == count
+
+    # Partial data.
+    cases = (
+        ({
+            "title": "digital libraries",
+            "ext_ids": {}
+        }, 5),
+        ({
+            "title": "unlikelytitle",
+            "ext_ids": {}
+        }, 0),
+        ({
+            "title": "Imminent dystopia",
+            "ext_ids": {}
+        }, 5),
+        ({
+            "title": "",
+            "contribs": [{
+                "raw_name": "Aristoteles"
+            }],
+            "ext_ids": {}
+        }, 5),
+        # ({
+        #     "title": "Letter",
+        #     "contribs": [{"raw_name": "Claudel"}],
+        #     "ext_ids": {}
+        # }, 1),
+        # ({
+        #     "title": "The Future of Digital Scholarship",
+        #     "contribs": [{
+        #         "raw_name": "Costantino Thanos"
+        #     }],
+        #     "ext_ids": {}
+        # }, 5),
+    )
+    for i, (doc, count) in enumerate(cases):
+        entity = entity_from_dict(doc, ReleaseEntity)
+        result = matcher.match(entity)
+        with caplog.at_level(logging.INFO):
+            logging.info("[{}] given title '{}', found {}, {}".format(i, entity.title, len(result),
+                                                                      [v.title for v in result]))
+        assert len(result) == count, doc
+
+
+def test_fuzzy_release_matcher_match_release_by_id(es_client, caplog):
+    matcher = FuzzyReleaseMatcher(es=es_client)
+    cases = (
+        ("wtv64ahbdzgwnan7rllwr3nurm", 1),
+        ("eqcgtpav3na5jh56o5vjsvb4ei", 1),
+    )
+    for i, (ident, count) in enumerate(cases):
+        entity = anything_to_entity(ident, ReleaseEntity)
+        result = matcher.match_release_by_id(entity)
+        assert len(result) == count
+
+
+def test_fuzzy_release_match_release_exact_title_exact_contrib(es_client, caplog):
+    matcher = FuzzyReleaseMatcher(es=es_client)
+    Case = collections.namedtuple("Case", "title date input expected")
+    cases = yaml_to_cases(
+        Case, "tests/files/fuzzy_release_match_release_exact_title_exact_contrib/*.yaml")
+    for i, c in enumerate(cases):
+        entity = entity_from_json(c.input, ReleaseEntity)
+        result = matcher.match_release_exact_title_exact_contrib(entity)
+        assert len(result) == c.expected, "[{}] {}".format(c.title, c.input)
+
+
+def test_fuzzy_release_match_release_exact_title_partial_contrib(es_client, caplog):
+    matcher = FuzzyReleaseMatcher(es=es_client)
+    Case = collections.namedtuple("Case", "title date input jaccard_index_threshold expected")
+    cases = yaml_to_cases(
+        Case, "tests/files/fuzzy_release_match_release_exact_title_partial_contrib/*.yaml")
+    for i, c in enumerate(cases):
+        entity = entity_from_json(c.input, ReleaseEntity)
+        result = matcher.match_release_exact_title_partial_contrib(entity)
+        assert len(result) == c.expected, "[{}] {}".format(c.title, c.input)