Add PolBlogs Dataset (#105)

* add PolBlogs

* add PolBlogs

* add tests

* add tests

* updated MLDatasets.jl

* add docstrings and tests

* add docstrings

* updated docs

* updated docs

* corrected index

* corrected spelling

* updated type

* updated type

* spelling corrected

* updated docstring

* updated docstring

* resolved renaming conficts

* updated docs

* minor changes

* resolved styling issues

Co-authored-by: Carlo Lucibello <[email protected]>

* renamed function

* updated docstrings

* updated type

Co-authored-by: Carlo Lucibello <[email protected]>

Author: soham-chitnis10

docs/src/datasets/graphs.md

@@ -10,6 +10,9 @@ CiteSeer.dataset
 Cora
 Cora.dataset
 OGBDataset
+PolBlogs
+PolBlogs.edge_index
+PolBlogs.labels
 PubMed
 PubMed.dataset
 TUDataset

src/MLDatasets.jl

@@ -59,7 +59,7 @@ include("planetoid.jl")
     include("CiteSeer/CiteSeer.jl")
 include("TUDataset/TUDataset.jl")
 include("OGBDataset/OGBDataset.jl")
-
+include("PolBlogs/PolBlogs.jl")
 
 function __init__()
     # initialize optional dependencies

src/PolBlogs/PolBlogs.jl

@@ -0,0 +1,76 @@
+export PolBlogs
+
+
+"""
+    PolBlogs
+ 
+The Political Blogs dataset from the [The Political Blogosphere and
+the 2004 US Election: Divided they Blog](https://dl.acm.org/doi/10.1145/1134271.1134277) paper.
+
+`PolBlogs` is a graph with 1,490 vertices (representing political blogs) and 19,025 edges (links between blogs).
+
+The links are automatically extracted from a crawl of the front page of the blog. 
+
+Each vertex receives a label indicating the political leaning of the blog: liberal or conservative.
+
+# Interface
+
+- [`PolBlogs.edge_index`](@ref)
+- [`PolBlogs.labels`](@ref)
+"""
+module PolBlogs
+
+using DataDeps
+using DelimitedFiles
+using ..MLDatasets: datafile
+
+export edge_index, labels
+
+const LINK = "https://netset.telecom-paris.fr/datasets/polblogs.tar.gz"
+const DEPNAME = "PolBlogs"
+const DATA = ["adjacency.csv", "labels.csv"]
+
+function __init__()
+    register(DataDep(DEPNAME,
+    """
+    Dataset : The $DEPNAME dataset
+    Website : $LINK
+    """,
+    LINK,
+    post_fetch_method = unpack
+    ))
+end
+
+"""
+    edge_index(; dir = nothing)
+
+Returns a 19025 x 2 matrix containing edge indices where first column as source node and second column as target node together they represent an edge
+
+```julia-repl
+using MLDatasets: PolBlogs
+adj = PolBlogs.edge_index()
+```
+"""
+function edge_index(; dir = nothing)
+    path = datafile(DEPNAME, DATA[1], dir)
+    adj = readdlm(path, ',')
+    return Matrix{Int64}(adj)
+end
+
+"""
+    labels(; dir = nothing)
+
+Returns a vector containing the 1490 labels.
+
+```julia-repl
+using MLDatasets: PolBlogs
+labels = PolBlogs.labels()
+```
+"""
+function labels(; dir=nothing)
+    path = datafile(DEPNAME, DATA[2], dir)
+    labels = readdlm(path, ',')
+    Matrix{Int64}(labels) |> vec
+end
+
+end #module

test/runtests.jl

@@ -27,6 +27,7 @@ dataset_tests = [
     "tst_cora.jl",
     "tst_pubmed.jl",
     "tst_tudataset.jl",
+    "tst_polblogs.jl",
 ]
 
 container_tests = [

test/tst_polblogs.jl

@@ -0,0 +1,12 @@
+data_dir = withenv("DATADEPS_ALWAY_ACCEPT"=>"true") do
+    datadep"PolBlogs"
+end
+
+@testset "PolBlogs" begin
+    adj = PolBlogs.edge_index()
+    labels = PolBlogs.labels()
+    @test adj isa Matrix{Int64}
+    @test labels isa Vector{Int64}
+    @test size(adj) == (19025,2)
+    @test size(labels) == (1490,)
+end