graphembed-rs 0.1.1__cp39-cp39-macosx_11_0_arm64.whl → 0.1.2__cp39-cp39-macosx_11_0_arm64.whl

graphembed_rs/__init__.py

@@ -1,5 +1,7 @@
-from .graphembed_rs import *
+from importlib import import_module as _imp
+from graphembed_rs import *      
 
-__doc__ = graphembed_rs.__doc__
-if hasattr(graphembed_rs, "__all__"):
-    __all__ = graphembed_rs.__all__
+# add the helper into the same namespace
+from .load_utils import *                   
+
+del _imp    

graphembed_rs/load_utils.py

@@ -0,0 +1,71 @@
+import numpy as np
+from pathlib import Path
+### install pymongo with pip install pymongo
+# -- pymongo is a BSON decoder, not a BSON encoder --
+from bson import decode_file_iter
+
+
+def _dtype_from_type_name(tname: str):
+    """Map the Rust type-name string to a NumPy dtype."""
+    if tname == "f32":
+        return np.float32
+    if tname == "f64":
+        return np.float64
+    if tname == "usize":                 # stored as i64 in BSON
+        return np.int64
+    raise ValueError(f"Unknown type_name {tname!r} in BSON header")
+
+
+def load_embedding_bson(path: str | Path, *, want_in: bool = False):
+    """
+    Parameters
+    ----------
+    path : str | pathlib.Path
+        File written by graphembed::io::bson_dump(...)
+    want_in : bool, default False
+        • False … always return the OUT/source embedding (shape = (n, d))  
+        • True  … additionally return the IN/target embedding when
+          the dump is *asymmetric* (tuple(out, in_)).  For symmetric
+          dumps the second item is None.
+
+    Returns
+    -------
+    np.ndarray            (sym. dump or want_in=False)
+    OR (out_emb, in_emb)  (want_in=True)
+    """
+    path = Path(path)
+    with path.open("rb") as fh:
+        docs = decode_file_iter(fh)
+
+        # -- header ----------------------------------------------------
+        header = next(docs)["header"]
+        n      = int(header["nbdata"])
+        d      = int(header["dimension"])
+        sym    = bool(header["symetric"])
+        dtype  = _dtype_from_type_name(header["type_name"])
+        out_emb = np.empty((n, d), dtype=dtype)
+        in_emb  = None if sym or not want_in else np.empty((n, d), dtype=dtype)
+
+        # --OUT part --------------------------------------------------
+        for _ in range(n):
+            doc = next(docs)
+            key, vec = next(iter(doc.items()))            # only 1 (key,val)
+            idx, tag = map(int, key.split(","))
+            assert tag == 0, f"expected tag 0, got {tag}"
+            out_emb[idx] = np.asarray(vec, dtype=dtype)
+
+        # -- IN part (if any) -----------------------------------------
+        if not sym:
+            for _ in range(n):
+                doc = next(docs)
+                key, vec = next(iter(doc.items()))
+                idx, tag = map(int, key.split(","))
+                assert tag == 1, f"expected tag 1, got {tag}"
+                if in_emb is not None:                    # want_in == True
+                    in_emb[idx] = np.asarray(vec, dtype=dtype)
+                # else: silently drop it
+
+        # -- optional indexation doc – skip for now --------------------
+        # (decode_file_iter stops automatically at EOF)
+
+    return (out_emb, in_emb) if want_in else out_emb

{graphembed_rs-0.1.1.dist-info → graphembed_rs-0.1.2.dist-info}/METADATA

@@ -1,6 +1,8 @@
 Metadata-Version: 2.4
 Name: graphembed_rs
-Version: 0.1.1
+Version: 0.1.2
+Requires-Dist: numpy>=2.2.5
+Requires-Dist: pymongo>=4.12.1
 License-File: LICENSE-MIT
 Summary: Python bindings for the high‑performance Rust graph/network embedding library graphembed
 Keywords: graph,embedding,hash
@@ -18,6 +20,9 @@ Project-URL: Source Code, https://github.com/jean-pierreBoth/graphembed
 ![](https://anaconda.org/bioconda/graphembed/badges/platforms.svg)
 [![install with conda](https://anaconda.org/bioconda/graphembed/badges/downloads.svg)](https://anaconda.org/bioconda/graphembed)
 
+<div align="center">
+  <img width="35%" src ="GraphEmbed_log.jpg">
+</div>
 
 # GraphEmbed: Efficient and Robust Network Embedding via High-Order Proximity Preservation or Recursive Sketching
 
@@ -30,13 +35,13 @@ This crate provides an executable and a library for embedding of directed or und
 
 ### Pre-built binaries on Linux
 ```bash
-wget https://gitlab.com/-/project/64961144/uploads/ea72ca007e9e4899e0c830e708f52939/graphembed_Linux_x86-64_v0.1.4.zip
-unzip graphembed_Linux_x86-64_v0.1.4.zip
+wget https://gitlab.com/-/project/64961144/uploads/9d7d0b038140cb67c584f01cd6dafac9/graphembed_Linux_x86-64_v0.1.6.zip
+unzip graphembed_Linux_x86-64_v0.1.6.zip
 chmod a+x ./graphembed
 ./graphembed -h
 ```
 
-### Bioconda on Linux
+### Bioconda on Linux/MacOS
 ```bash
 conda install -c conda-forge -c bioconda graphembed
 ```
@@ -49,7 +54,7 @@ brew install graphembed
 ```
 
 
-### In Python (Please install python first)
+### In Python (Please install python>=3.9 first)
 ```bash
 pip install graphembed_rs
 
@@ -57,7 +62,7 @@ pip install graphembed_rs
 git clone https://gitlab.com/Jianshu_Zhao/graphembed
 cd graphembed
 pip install maturin
-### note: for macOS, you need to change the line "features = ["pyo3/extension-module", "intel-mkl-static", "simdeez_f"]" in pyporject.toml to "features = ["pyo3/extension-module","openblas-system","stdsimd"]"
+### note: for macOS, you need to change the line "features = ["pyo3/extension-module", "intel-mkl-static", "simdeez_f"]" in pyporject.toml to "features = ["pyo3/extension-module","openblas-system","stdsimd"]", you also need to Install OpenBLAS and add to system library path via Homebrew
 maturin develop --release
 
 #### Prepare some data
@@ -66,15 +71,25 @@ wget https://gitlab.com/-/project/64961144/uploads/4e341383d62d86d1dd66e668e91b2
 
 ```python
 import os
-os.environ["RUST_LOG"] = "graphembed=info"
-import graphembed as ge
+os.environ["RUST_LOG"] = "info"
+import graphembed_rs.graphembed_rs as ge
+import graphembed_rs.load_utils as ge_utils
 help(ge)
+help(ge_utils)
 ### HOPE
-ge.embed_hope_rank("BlogCatalog.txt", target_rank=128, nbiter=4)
+ge.embed_hope_rank("BlogCatalog.txt", target_rank=128, nbiter=4,output="embedding_output")
+out_vectors=ge_utils.load_embedding_bson("embedding_output.bson")
+print("OUT embedding shape :", out_vectors.shape)
+print("first OUT vector    :", out_vectors[0])
 
 ### Sketching
 ### sketching only
 ge.embed_sketching("BlogCatalog.txt", decay=0.3, dim=128, nbiter=5, symetric=True, output="embedding_output")
+out_vectors=ge_utils.load_embedding_bson("embedding_output.bson")
+print("OUT embedding shape :", out_vectors.shape)
+print("first OUT vector    :", out_vectors[0])
+
+
 ### validate accuracy
 auc_scores = ge.validate_sketching("BlogCatalog.txt",decay=0.3, dim=128, nbiter=3, nbpass=1, skip_frac=0.2,symetric=True, centric=True)
 print("Standard AUC per pass:", auc_scores)

graphembed_rs-0.1.2.dist-info/RECORD

@@ -0,0 +1,7 @@
+graphembed_rs-0.1.2.dist-info/METADATA,sha256=WEvYJP6qVNPFlFxYzzq-n3XX6L__a2ukxgYkVt_WnYo,11532
+graphembed_rs-0.1.2.dist-info/WHEEL,sha256=OWa2DoOWNAtKx8o8X2SfzZNCVYa4xySMUeG2PSXKaJ4,102
+graphembed_rs-0.1.2.dist-info/licenses/LICENSE-MIT,sha256=ndZ12D28O4UkfOeoa6HP9E7IKyYG4iH79iQ6WiLs9bc,1077
+graphembed_rs/load_utils.py,sha256=fXd4-5OxdymSkb4kx39XeW_P3nEDdAuQ7Oe1D6PKDWE,2719
+graphembed_rs/__init__.py,sha256=s8WorNcMnn15CfjctNaLnXGnwQ9RlPyCR0WJQxepTTc,179
+graphembed_rs/graphembed_rs.cpython-39-darwin.so,sha256=bnNJQ9TeEI6GVmt9GP_FJUtfH9enrlcYdvjbXfY5nEw,5265888
+graphembed_rs-0.1.2.dist-info/RECORD,,

graphembed_rs/__init__.pyi

@@ -1,79 +0,0 @@
-from typing import Optional
-
-# ----------  Embedding  ----------
-def embed_hope_rank(
-    csv: str,
-    symetric: bool,
-    target_rank: int,
-    nbiter: int,
-    output: Optional[str] = None,
-) -> None: ...
-def embed_hope_precision(
-    csv: str,
-    symetric: bool,
-    epsil: float,
-    maxrank: int,
-    blockiter: int,
-    output: Optional[str] = None,
-) -> None: ...
-def embed_sketching(
-    csv: str,
-    symetric: bool,
-    decay: float,
-    dim: int,
-    nbiter: int,
-    output: Optional[str] = None,
-) -> None: ...
-
-# ----------  Validation (returns mean AUC) ----------
-def validate_hope_rank(
-    csv: str,
-    symetric: bool,
-    target_rank: int,
-    nbiter: int,
-    nbpass: int = 10,
-    skip_frac: float = 0.1,
-    centric: bool = False,
-) -> float: ...
-def validate_hope_precision(
-    csv: str,
-    symetric: bool,
-    epsil: float,
-    maxrank: int,
-    blockiter: int,
-    nbpass: int = 10,
-    skip_frac: float = 0.1,
-    centric: bool = False,
-) -> float: ...
-def validate_sketching(
-    csv: str,
-    symetric: bool,
-    decay: float,
-    dim: int,
-    nbiter: int,
-    nbpass: int = 10,
-    skip_frac: float = 0.1,
-    centric: bool = False,
-) -> float: ...
-
-# ---------- VCMPR (precision/recall curves) ----------
-def estimate_vcmpr_hope_rank(
-    csv: str,
-    symetric: bool,
-    target_rank: int,
-    nbiter: int,
-    nbpass: int = 2,
-    topk: int = 10,
-    skip_frac: float = 0.1,
-) -> None: ...
-def estimate_vcmpr_sketching(
-    csv: str,
-    symetric: bool,
-    decay: float,
-    dim: int,
-    nbiter: int,
-    nbpass: int = 2,
-    topk: int = 10,
-    skip_frac: float = 0.1,
-) -> None: ...
-

graphembed_rs-0.1.1.dist-info/RECORD

@@ -1,8 +0,0 @@
-graphembed_rs-0.1.1.dist-info/METADATA,sha256=uW8W9UwifiE6FViIKpp1I7d5e6ygVuo6AA2O72MogLU,10890
-graphembed_rs-0.1.1.dist-info/WHEEL,sha256=OWa2DoOWNAtKx8o8X2SfzZNCVYa4xySMUeG2PSXKaJ4,102
-graphembed_rs-0.1.1.dist-info/licenses/LICENSE-MIT,sha256=ndZ12D28O4UkfOeoa6HP9E7IKyYG4iH79iQ6WiLs9bc,1077
-graphembed_rs/__init__.py,sha256=R2D0If_-sN__21LBYNod0CNgVo2dCd2RqM11AStM3X0,135
-graphembed_rs/__init__.pyi,sha256=3_KBFG4g9akylo32CHlm9bZStcLwxIY2X4si21ilD3w,1626
-graphembed_rs/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-graphembed_rs/graphembed_rs.cpython-39-darwin.so,sha256=uHq1hQqyDyoMhlm2Uz2XN0oDXK8omVehmCunu9ABg9w,5106880
-graphembed_rs-0.1.1.dist-info/RECORD,,