black

Author: fgregg

dedupe/__init__.py

@@ -1,7 +1,15 @@
-#!/usr/bin/python
-# -*- coding: utf-8 -*-
-from pkgutil import extend_path
-
-__path__ = extend_path(__path__, __name__)
-
-from dedupe._init import *  # noqa
+from dedupe.api import (  # noqa: F401
+    Dedupe,
+    Gazetteer,
+    RecordLink,
+    StaticDedupe,
+    StaticGazetteer,
+    StaticRecordLink,
+)
+from dedupe.convenience import (  # noqa: F401
+    canonicalize,
+    console_label,
+    training_data_dedupe,
+    training_data_link,
+)
+from dedupe.serializer import read_training, write_training  # noqa: F401

dedupe/_init.py

@@ -0,0 +1,111 @@
+import functools
+import warnings
+from typing import Any, Iterable, Mapping, Sequence
+
+from ._typing import Cover
+from .predicates import Predicate
+
+Partial = tuple[Predicate, ...]
+
+
+def _reachable(dupe_cover: Mapping[Any, frozenset[int]]) -> int:
+    return len(frozenset.union(*dupe_cover.values())) if dupe_cover else 0
+
+
+def _remove_dominated(coverage: Cover, dominator: Predicate) -> Cover:
+    dominant_cover = coverage[dominator]
+
+    return {
+        pred: cover
+        for pred, cover in coverage.items()
+        if not (dominator.cover_count <= pred.cover_count and dominant_cover >= cover)
+    }
+
+
+def _uncovered_by(
+    coverage: Mapping[Any, frozenset[int]], covered: frozenset[int]
+) -> dict[Any, frozenset[int]]:
+    remaining = {}
+    for predicate, uncovered in coverage.items():
+        still_uncovered = uncovered - covered
+        if still_uncovered:
+            remaining[predicate] = still_uncovered
+
+    return remaining
+
+
+def _order_by(
+    candidates: Mapping[Predicate, Sequence[Any]], p: Predicate
+) -> tuple[int, float]:
+    return (len(candidates[p]), -p.cover_count)
+
+
+def _score(partial: Iterable[Predicate]) -> float:
+    return sum(p.cover_count for p in partial)
+
+
+def _suppress_recursion_wrapper(func):
+    def wrapper(*args, **kwargs):
+        try:
+            return func(*args, **kwargs)
+        except RecursionError:
+            warnings.warn("Recursion limit eached while searching for predicates")
+
+    return wrapper
+
+
+def search(candidates, target: int, max_calls: int) -> Partial:
+    calls = max_calls
+
+    cheapest_score = float("inf")
+    cheapest: Partial = ()
+
+    original_cover = candidates.copy()
+
+    def _covered(partial: Partial) -> int:
+        return (
+            len(frozenset.union(*(original_cover[p] for p in partial)))
+            if partial
+            else 0
+        )
+
+    @_suppress_recursion_wrapper
+    def walk(candidates: Cover, partial: Partial = ()) -> None:
+        nonlocal calls
+        nonlocal cheapest
+        nonlocal cheapest_score
+
+        if calls <= 0:
+            return
+
+        calls -= 1
+
+        covered = _covered(partial)
+        score = _score(partial)
+
+        if covered < target:
+            window = cheapest_score - score
+            candidates = {
+                p: cover for p, cover in candidates.items() if p.cover_count < window
+            }
+
+            reachable = _reachable(candidates) + covered
+
+            if candidates and reachable >= target:
+                order_by = functools.partial(_order_by, candidates)
+                best = max(candidates, key=order_by)
+
+                remaining = _uncovered_by(candidates, candidates[best])
+                walk(remaining, partial + (best,))
+                del remaining
+
+                reduced = _remove_dominated(candidates, best)
+                walk(reduced, partial)
+                del reduced
+
+        elif score < cheapest_score:
+            cheapest = partial
+            cheapest_score = score
+
+    walk(candidates)
+    return cheapest

dedupe/branch_and_bound.py

@@ -10,10 +10,10 @@
 from typing import TYPE_CHECKING, overload
 from warnings import warn
 
-from . import blocking
+from . import blocking, branch_and_bound
 
 if TYPE_CHECKING:
-    from typing import Any, Iterable, Mapping, Sequence
+    from typing import Iterable, Sequence
 
     from ._typing import (
         ComparisonCover,
@@ -75,8 +75,7 @@ def learn(
         else:
             raise ValueError("candidate_type is not valid")
 
-        searcher = BranchBound(target_cover, 2500)
-        final_predicates = searcher.search(candidate_cover)
+        final_predicates = branch_and_bound.search(candidate_cover, target_cover, 2500)
 
         logger.info("Final predicate set:")
         for predicate in final_predicates:
@@ -329,113 +328,6 @@ def coveredPairs(self, blocker, records_1, records_2):
         return pair_cover
 
 
-class BranchBound(object):
-    def __init__(self, target: int, max_calls: int) -> None:
-        self.target: int = target
-        self.calls: int = max_calls
-
-        self.cheapest_score: float = float("inf")
-        self.original_cover: Cover = {}
-        self.cheapest: tuple[Predicate, ...] = ()
-
-    def search(
-        self, candidates: Cover, partial: tuple[Predicate, ...] = ()
-    ) -> tuple[Predicate, ...]:
-        if self.calls <= 0:
-            return self.cheapest
-
-        if not self.original_cover:
-            self.original_cover = candidates.copy()
-
-        self.calls -= 1
-
-        covered = self.covered(partial)
-        score = self.score(partial)
-
-        if covered >= self.target:
-            if score < self.cheapest_score:
-                self.cheapest = partial
-                self.cheapest_score = score
-
-        else:
-            window = self.cheapest_score - score
-
-            candidates = {
-                p: cover for p, cover in candidates.items() if p.cover_count < window
-            }
-
-            reachable = self.reachable(candidates) + covered
-
-            if candidates and reachable >= self.target:
-                order_by = functools.partial(self.order_by, candidates)
-
-                best = max(candidates, key=order_by)
-
-                remaining = self.uncovered_by(candidates, candidates[best])
-                try:
-                    self.search(remaining, partial + (best,))
-                except RecursionError:
-                    return self.cheapest
-
-                del remaining
-
-                reduced = self.remove_dominated(candidates, best)
-
-                try:
-                    self.search(reduced, partial)
-                except RecursionError:
-                    return self.cheapest
-
-                del reduced
-
-        return self.cheapest
-
-    @staticmethod
-    def order_by(
-        candidates: Mapping[Predicate, Sequence[Any]], p: Predicate
-    ) -> tuple[int, float]:
-        return (len(candidates[p]), -p.cover_count)
-
-    @staticmethod
-    def score(partial: Iterable[Predicate]) -> float:
-        return sum(p.cover_count for p in partial)
-
-    def covered(self, partial: tuple[Predicate, ...]) -> int:
-        if partial:
-            return len(frozenset.union(*(self.original_cover[p] for p in partial)))
-        else:
-            return 0
-
-    @staticmethod
-    def reachable(dupe_cover: Mapping[Any, frozenset[int]]) -> int:
-        if dupe_cover:
-            return len(frozenset.union(*dupe_cover.values()))
-        else:
-            return 0
-
-    @staticmethod
-    def remove_dominated(coverage: Cover, dominator: Predicate) -> Cover:
-        dominant_cover = coverage[dominator]
-
-        for pred, cover in coverage.copy().items():
-            if dominator.cover_count <= pred.cover_count and dominant_cover >= cover:
-                del coverage[pred]
-
-        return coverage
-
-    @staticmethod
-    def uncovered_by(
-        coverage: Mapping[Any, frozenset[int]], covered: frozenset[int]
-    ) -> dict[Any, frozenset[int]]:
-        remaining = {}
-        for predicate, uncovered in coverage.items():
-            still_uncovered = uncovered - covered
-            if still_uncovered:
-                remaining[predicate] = still_uncovered
-
-        return remaining
-
-
 class InfiniteSet(object):
     def __and__(self, item):
         return item

dedupe/training.py

@@ -28,7 +28,6 @@ dependencies = [
     "scikit-learn",
     "affinegap>=1.3",
     "categorical-distance>=1.9",
-    "dedupe-variable-datetime",
     "numpy>=1.20",
     "doublemetaphone",
     "highered>=0.2.0",