Merge pull request #14 from jbrea/dev

Remove ScientificTypes & Improve error messages

Project.toml

@@ -1,21 +1,19 @@
 name = "OpenML"
 uuid = "8b6db2d4-7670-4922-a472-f9537c81ab66"
 authors = ["Diego Arenas <[email protected]>", "Anthony D. Blaom <[email protected]>"]
-version = "0.1.1"
+version = "0.2.0"
 
 [deps]
 ARFFFiles = "da404889-ca92-49ff-9e8b-0aa6b4d38dc8"
 HTTP = "cd3eb016-35fb-5094-929b-558a96fad6f3"
 JSON = "682c06a0-de6a-54ab-a142-c8b1cf79cde6"
 Markdown = "d6f4376e-aef5-505a-96c1-9c027394607a"
 Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
-ScientificTypes = "321657f4-b219-11e9-178b-2701a2544e81"
 
 [compat]
 ARFFFiles = "1.3"
 HTTP = "0.8, 0.9"
 JSON = "0.21"
-ScientificTypes = "2"
 julia = "1"
 
 [extras]

src/OpenML.jl

@@ -1,9 +1,8 @@
-module OpenML 
+module OpenML
 
 using HTTP
 using JSON
 import ARFFFiles
-import ScientificTypes: Continuous, Count, Textual, Multiclass, coerce, autotype
 using Markdown
 if VERSION > v"1.3.0"
     using Pkg.Artifacts

src/data.jl

@@ -28,34 +28,35 @@ function load_Dataset_Description(id::Int; api_key::String="")
             println("No access granted. This dataset is not shared with you.")
         end
     catch e
-        println("Error occurred : $e")
+        println("Error occurred. Check if there exists a dataset with id $id.")
+        println("See e.g. OpenML.list_datasets()\n")
+        println(e)
         return nothing
     end
     return nothing
 end
 
 """
-    OpenML.load(id; parser = :arff)
+    OpenML.load(id)
 
 Load the OpenML dataset with specified `id`, from those listed by
 [`list_datasets`](@ref) or on the [OpenML site](https://www.openml.org/search?type=data).
-With `parser = :arff` (default) the ARFFFiles.jl parser is used.
-With `parser = :auto` the output of the ARFFFiles parser is coerced to
-automatically detected scientific types.
 
-Datasets are saved as julia artifacts so that they persist locally once loaded. 
+Datasets are saved as julia artifacts so that they persist locally once loaded.
 
 Returns a table.
 
 # Examples
 
 ```julia
 using DataFrames
-table = OpenML.load(61);
-df = DataFrame(table);
+table = OpenML.load(61)
+df = DataFrame(table) # transform to a DataFrame
+using ScientificTypes
+df2 = coerce(df, autotype(df)) # corce to automatically detected scientific types
 ```
 """
-function load(id::Int; parser = :arff)
+function load(id::Int)
     if VERSION > v"1.3.0"
         dir = first(Artifacts.artifacts_dirs())
         toml = joinpath(dir, "OpenMLArtifacts.toml")
@@ -73,12 +74,7 @@ function load(id::Int; parser = :arff)
         filename = tempname()
         download(url, filename)
     end
-    data = ARFFFiles.load(filename)
-    if parser == :auto
-        return coerce(data, autotype(data))
-    else
-        return data
-    end
+    ARFFFiles.load(filename)
 end
 
 
@@ -321,7 +317,14 @@ julia> OpenML.describe_dataset(6)
   cited above for more details.
 ```
 """
-describe_dataset(id) =  Markdown.parse(load_Dataset_Description(id)["data_set_description"]["description"])
+function describe_dataset(id)
+    description = load_Dataset_Description(id)["data_set_description"]["description"]
+    if isa(description, AbstractString)
+        Markdown.parse(description)
+    else
+        "No description found."
+    end
+end
 
 # Flow API
 

test/data.jl

@@ -41,12 +41,12 @@ end
 end
 
 if VERSION > v"1.3.0"
-    using Pkg.Artifacts
+    using Pkg
     @testset "artifacts" begin
-        dir = first(Artifacts.artifacts_dirs())
+        dir = first(Pkg.Artifacts.artifacts_dirs())
         toml = joinpath(dir, "OpenMLArtifacts.toml")
-        hash = artifact_hash("61", toml)
-        @test artifact_exists(hash)
+        hash = Pkg.Artifacts.artifact_hash("61", toml)
+        @test Pkg.Artifacts.artifact_exists(hash)
     end
 end