add MatMiner datasets

docs/src/datasets/misc.md

@@ -11,6 +11,7 @@ Pages = ["misc.md"]
 ```@docs
 BostonHousing
 Iris
+MatMiner
 Mutagenesis
 Titanic
 ```

src/MLDatasets.jl

@@ -66,6 +66,8 @@ include("datasets/misc/boston_housing.jl")
 export BostonHousing
 include("datasets/misc/iris.jl")
 export Iris
+include("datasets/misc/matminer.jl")
+export MatMiner
 include("datasets/misc/mutagenesis.jl")
 export Mutagenesis
 include("datasets/misc/titanic.jl")
@@ -150,6 +152,7 @@ function __init__()
 
     # misc
     __init__iris()
+    __init__matminer()
     __init__mutagenesis()
 
     #text

src/datasets/misc/matminer.jl

@@ -0,0 +1,68 @@
+function __init__matminer()
+    DEPNAME = "MatMiner"
+    DOCS = "https://hackingmaterials.lbl.gov/matminer/"
+
+    register(ManualDataDep(
+        DEPNAME,
+        """
+        Datahub: $DEPNAME.
+        Website: $DOCS
+        """,
+    ))
+end
+
+"""
+    MatMiner(name; dir=nothing)
+
+
+# Examples
+
+```julia
+data = MatMiner("steels")
+```
+"""
+struct MatMiner <: AbstractDataset
+    name::String
+    metadata
+    data
+end
+
+function MatMiner(name::String; dir=nothing)
+    jsondata, metadata, dir = matminer_getdata(name; dir)
+    data = JSON3.copy(jsondata)
+    metadata = JSON3.copy(metadata)
+    return MatMiner(name, metadata, data)
+end
+
+function matminer_metadata(mmdir)
+    LINK = "https://raw.githubusercontent.com/hackingmaterials/matminer/main/matminer/datasets/dataset_metadata.json"
+    path = joinpath(mmdir, "dataset_metadata.json")
+    if !isfile(path)
+        DataDeps.fetch_default(LINK, mmdir)
+    end
+    @assert isfile(path)
+    return read_json(path)
+end
+
+function matminer_getdata(name; dir = nothing)
+    create_default_dir("MatMiner")
+    mmdir = isnothing(dir) ? datadep"MatMiner" : dir
+    metadata = matminer_metadata(mmdir)
+    @assert name in keys(metadata) "Dataset $name not found among $(string.(keys(metadata)))"
+    if !isdir(joinpath(mmdir, name))
+        mkdir(joinpath(mmdir, name))
+    end
+    dir = joinpath(mmdir, name) 
+    url = metadata[name]["url"]
+    @assert metadata[name]["file_type"] == "json.gz"
+    filepath = joinpath(dir, name * ".json")
+    if !isfile(filepath)
+        DataDeps.fetch_default(url, dir)
+        currdir = pwd()
+        cd(dir) # Needed since `unpack` extracts in working dir
+        DataDeps.unpack(filepath * ".gz")
+        cd(currdir)
+    end
+    jsondata = read_json(filepath)
+    return jsondata, metadata, dir
+end