switch housing dataset to wine

[Ivanidzo4ka]

This commit replaces housing dataset to wine dataset which we download during build from external source

[Ivanidzo4ka]

address #3

[eerhardt]

Can you do me a favor and test this with 2 or 3 datasets downloaded from the internet? (I guess it can be any file, it doesn't have to be a .csv file) I just want to ensure this works with more than 1 file, and it works correctly when doing it a 2nd time that the data sets aren't downloaded on subsequent builds.

That way the next person who needs to add a dataset isn't bit by a bug here.

[Ivanidzo4ka]

<ItemGroup> <TestFile Include="$(MSBuildThisFileDirectory)/test/data/external/winequality-white.csv" Url="https://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-white.csv"> <DestinationFile>$(MSBuildThisFileDirectory)test/data/external/winequality-white.csv</DestinationFile> </TestFile> <TestFile Include="$(MSBuildThisFileDirectory)/test/data/external/image.png" Url="https://www.google.com/logos/doodles/2018/tamara-de-lempickas-120th-birthday-4614326680813568-l.png"> <DestinationFile>$(MSBuildThisFileDirectory)test/data/external/image.png</DestinationFile> </TestFile> <TestFile Include="$(MSBuildThisFileDirectory)/test/data/external/logo.png" Url="https://www.baidu.com/img/bd_logo1.png"> <DestinationFile>$(MSBuildThisFileDirectory)test/data/external/logo.png</DestinationFile> </TestFile> </ItemGroup>
test it with this. tried different combinations, they all work
Is there are any way to format it properly?

[eerhardt]

Awesome! nice to hear it works for many items.

As for the formatting, if you use triple back-ticks ``` at the start of a line, markdown will show exactly what you type until the next set of triple back-ticks. This is useful for displaying code. You can even put a code hint after the triple back-ticks XML, and markdown will syntax highlight. So if I take your comment above it looks like this:

<ItemGroup>
    <TestFile Include="$(MSBuildThisFileDirectory)/test/data/external/winequality-white.csv" Url="https://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-white.csv">
         <DestinationFile>$(MSBuildThisFileDirectory)test/data/external/winequality-white.csv</DestinationFile>
    </TestFile>
    <TestFile Include="$(MSBuildThisFileDirectory)/test/data/external/image.png" Url="https://www.google.com/logos/doodles/2018/tamara-de-lempickas-120th-birthday-4614326680813568-l.png">
<DestinationFile>$(MSBuildThisFileDirectory)test/data/external/image.png</DestinationFile>
</TestFile>
<TestFile Include="$(MSBuildThisFileDirectory)/test/data/external/logo.png" Url="https://www.baidu.com/img/bd_logo1.png">         <DestinationFile>$(MSBuildThisFileDirectory)test/data/external/logo.png</DestinationFile>
    </TestFile>
  </ItemGroup>

See https://guides.github.com/features/mastering-markdown/#GitHub-flavored-markdown for more info.

[eerhardt]

I see a few more references to housing.txt in the test code:

F:\git\machinelearning\test\Microsoft.ML.Core.Tests\UnitTests\TestCSharpApi.cs(275):            var dataPath = GetDataPath(@"housing.txt");
  F:\git\machinelearning\test\Microsoft.ML.Core.Tests\UnitTests\TestEntryPoints.cs(861):            var dataPath = GetDataPath("housing.txt");
  F:\git\machinelearning\test\Microsoft.ML.Core.Tests\UnitTests\TestEntryPoints.cs(893):            TestEntryPointRoutine("housing.txt", "Trainers.StochasticDualCoordinateAscentRegressor");
  F:\git\machinelearning\test\Microsoft.ML.Core.Tests\UnitTests\TestEntryPoints.cs(967):            TestEntryPointRoutine("housing.txt", "Trainers.PoissonRegressor");
  F:\git\machinelearning\test\Microsoft.ML.TestFramework\Datasets.cs(151):            trainFilename = "housing.txt",
  F:\git\machinelearning\test\Microsoft.ML.TestFramework\Datasets.cs(152):            testFilename = "housing.txt"
  F:\git\machinelearning\test\Microsoft.ML.TestFramework\TestCommandBase.cs(1496):                string pathData = GetDataPath(@"..\Housing (regression)", "housing.txt");
  F:\git\machinelearning\test\Microsoft.ML.TestFramework\TestCommandBase.cs(1536):            string pathData = GetDataPath(@"..\Housing (regression)", "housing.txt");
  F:\git\machinelearning\test\Microsoft.ML.TestFramework\TestCommandBase.cs(1608):            string pathData = GetDataPath(@"..\Housing (regression)", "housing.txt");
  F:\git\machinelearning\test\Microsoft.ML.TestFramework\TestCommandBase.cs(1649):                string pathData = GetDataPath(@"..\Housing (regression)", "housing.txt");

Can you update them all?

[danmoseley]

Nit, I believe the attribute and subelement formats are equivalent, you might want to pick one or the other. #Resolved

[Ivanidzo4ka]

I can't move Include to subelement, but I can move Url to it. having everything as attribute is looks too long

---

In reply to: 189065564 [](ancestors = 189065564)

[eerhardt]

I'm not a requesting change here, just giving some information for future reference.

The Include attribute on an MSBuild Item is "special". It has to be an XML attribute. Before MSBuild v15, all "custom" metadata (like your Url attribute) could NOT be put as an XML attribute - instead it had to be a sub-element. In v15, they changed it so custom metadata now can be in either an attribute or sub-element.

A format I like and use regularly is the following:

    <TestFile Include="$(MSBuildThisFileDirectory)/test/data/external/winequality-white.csv"
              Url="https://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-white.csv"
              DestinationFile="$(MSBuildThisFileDirectory)test/data/external/winequality-white.csv" />

I think it is compact, reads easy, and is easy to maintain/update.

[Ivanidzo4ka]

We don't run any of these tests. #78 Anirudh has PR to enable them, and it should be resolved in that PR. (Either by Anirudh, or me)

---

In reply to: 389680637 [](ancestors = 389680637)

[Ivanidzo4ka]

@dotnet-bot test Windows_NT Debug

[eerhardt]

[Ivanidzo4ka]

Anyone other than Eric can look on this? @glebuk

[glebuk]

        using (var loader = new BinaryLoader(Env, new BinaryLoader.Arguments(), confusionMatrixPath))

Missing TestCSarpAPI?
#Closed

---

Refers to: test/Microsoft.ML.Core.Tests/UnitTests/TestEntryPoints.cs:1037 in 45478c2. [](commit_id = 45478c2, deletion_comment = False)

[glebuk]

@"external/winequality-white.csv [](start = 39, length = 32)

Extract dataset info into static dataset classes so that no need to repeat paths and schemas. Ideally you can even just return a new loader for each dataset. #Closed

[glebuk]