bmp 0.2__cp37-none-win_amd64.whl

bmp/__init__.py

@@ -0,0 +1 @@
+from bmp._bmp import ciff2bmp, search, Searcher, InvertedIndexer, Indexer

bmp/pyterrier.py

@@ -0,0 +1,79 @@
+import numpy as np
+from pathlib import Path
+import pyterrier as pt
+import pyterrier_alpha as pta
+from bmp import Indexer, Searcher
+
+class BmpIndex(pta.Artifact, pt.Indexer):
+    ARTIFACT_TYPE = 'sparse_index'
+    ARTIFACT_FORMAT = 'bmp'
+    def __init__(self, path: str, *, verbose: bool = True):
+        super().__init__(path)
+        self.verbose = verbose
+        self._searcher = None
+    def built(self) -> bool:
+        return Path(self.path).exists()
+    def indexer(self, bsize=32, compress_range=False, scale_float=100.):
+        return BmpIndexer(self, bsize=bsize, compress_range=compress_range, scale_float=scale_float)
+    def index(self, inp):
+        return self.indexer().index(inp)
+    def retriever(self):
+        return BmpRetriever(self)
+    def transform(self, inp):
+        return self.retriever()(inp)
+    def load_into_memory(self):
+        if self._searcher is None:
+            self._searcher = Searcher(str(self.path/'index.bmp'))
+        return self._searcher
+    def __enter__(self):
+        return self
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        self.close()
+    def close(self):
+        self._searcher = None
+
+class BmpIndexer(pt.Indexer):
+    def __init__(self, bmp_index, bsize=32, compress_range=False, scale_float=100.):
+        self.bmp_index = bmp_index
+        self.bsize = bsize
+        self.compress_range = compress_range
+        self.scale_float = scale_float
+    def index(self, inp):
+        assert not self.bmp_index.built()
+        with pta.ArtifactBuilder(self.bmp_index) as builder:
+            indexer = Indexer(str(self.bmp_index.path/'index.bmp'), bsize=self.bsize, compress_range=self.compress_range)
+            count = 0
+            for doc in inp:
+                vector = doc['toks']
+                if len(vector) > 0 and isinstance(next(iter(vector.values())), float):
+                    vector = {k: int(v * self.scale_float) for k, v in vector.items()}
+                indexer.add_document(doc['docno'], vector)
+                count += 1
+            indexer.finish()
+            builder.metadata['bsize'] = self.bsize
+            builder.metadata['compress_range'] = self.compress_range
+            builder.metadata['scale_float'] = self.scale_float
+            builder.metadata['num_docs'] = count
+        return self
+
+class BmpRetriever(pt.Indexer):
+    def __init__(self, bmp_index, *, num_results=1000, alpha=1.0, beta=1.0):
+        self.bmp_index = bmp_index
+        self.num_results = num_results
+        self.alpha = alpha
+        self.beta = beta
+    def transform(self, inp):
+        pta.validate.query_frame(inp, extra_columns=['query_toks'])
+        searcher = self.bmp_index.load_into_memory()
+        res = pta.DataFrameBuilder(['docno', 'score', 'rank'])
+        for toks in inp['query_toks']:
+            docnos, scores = searcher.search(toks, k=self.num_results, alpha=self.alpha, beta=self.beta)
+            res.extend({
+                'docno': docnos,
+                'score': scores,
+                'rank': np.arange(len(scores))
+            })
+        return res.to_df(inp)
+    def fuse_rank_cutoff(self, k):
+        if self.num_results > k:
+            return BmpRetriever(self.bmp_index, num_results=k, alpha=self.apha, beta=self.beta)

bmp-0.2.dist-info/METADATA

@@ -0,0 +1,71 @@
+Metadata-Version: 2.1
+Name: bmp
+Version: 0.2
+Classifier: Programming Language :: Rust
+Classifier: Programming Language :: Python :: Implementation :: CPython
+Classifier: Programming Language :: Python :: Implementation :: PyPy
+Requires-Dist: python-terrier; extra == 'pyterrier'
+Requires-Dist: pyterrier-alpha; extra == 'pyterrier'
+Provides-Extra: pyterrier
+Requires-Python: >=3.7
+Description-Content-Type: text/markdown; charset=UTF-8; variant=GFM
+
+
+<p align="center">
+    <img width="100px" src="../img/logo.jpg" />
+    <img width="100px" src="../img/plus.png" />
+    <img width="100px" src="../img/python.png" />
+    <h1 align="center">BMP</h1>
+</p>
+
+## Install
+
+From PyPi:
+
+```
+pip install bmp
+```
+
+From Source:
+
+```
+pip install maturin
+maturin build -r
+pip install target/wheels/*.whl
+```
+
+## Usage
+
+### Index from CIFF
+
+```python
+from bmp import ciff2bmp
+ciff2bmp(ciff_file="/path/to/ciff", output="/path/to/index", bsize=32, compress_range=False)
+```
+
+### Index with Python
+
+```python
+from bmp import Indexer
+indexer = Indexer('/path/to/index', bsize=32, compress_range=False)
+indexer.add_document('doc1', {'a': 1, 'b': 5, 'c': 8}) # docid, vector
+indexer.add_document('doc2', {'a': 2, 'c': 1, 'd': 8, 'f': 2})
+...
+indexer.finish()
+```
+
+#### Search
+
+```python
+from bmp import search, Searcher
+
+# batch operation
+results = search(index="/path/to/index", queries="/path/to/queries", k=10, alpha=1.0, beta=1.0)
+# -> str (TREC run file)
+
+# query-by-query operation
+searcher = Searcher("/path/to/index") # loads index into memory once
+searcher.search({'tok1': 5.3, 'tok2': 1.1}, k=10, alpha=1.0, beta=1.0)
+# -> Tuple[List[str], List[float]] (doc IDs, scores) for this query
+```
+

bmp-0.2.dist-info/RECORD

@@ -0,0 +1,7 @@
+bmp-0.2.dist-info/METADATA,sha256=XCTAvc7nosH_ZAG5CPKw5onCbUydOMXASPYFPqOVHk0,1739
+bmp-0.2.dist-info/WHEEL,sha256=5EEVb2qtUBAftJIn0bFfpziowjNUtiDBo-e9pkgBCo8,96
+bmp-0.2.dist-info/entry_points.txt,sha256=7T9_5X3HnKORahbEc7zjcrSMKd4gluc-v7sXmgx7BMs,61
+bmp/pyterrier.py,sha256=JRRSGp9eXcnd0lvU3injJ3nUZENThdf4CpPJzWEiiR0,3317
+bmp/__init__.py,sha256=MlXKBU0lt2uGHecHHOW-ndvEOqjBEFBfqsrzAK-uNM8,75
+bmp/_bmp.pyd,sha256=MVSrwOrPVGw63d2wYuUplFnW2CfSKRa1_hf1XuHWpeA,1471488
+bmp-0.2.dist-info/RECORD,,

bmp-0.2.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: maturin (0.14.17)
+Root-Is-Purelib: false
+Tag: cp37-none-win_amd64

bmp-0.2.dist-info/entry_points.txt

@@ -0,0 +1,2 @@
+[pyterrier.artifact]
+sparse_index.bmp=bmp.pyterrier:BmpIndex