public void TestIndexCreateAndDelete()
{
// Should be one active index.
DataFrame indexes = _hyperspace.Indexes();
Assert.Equal(1, indexes.Count());
Assert.Equal(_sampleIndexName, indexes.SelectExpr("name").First()[0]);
Assert.Equal(States.Active, indexes.SelectExpr("state").First()[0]);
// Delete the index then verify it has been deleted.
_hyperspace.DeleteIndex(_sampleIndexName);
indexes = _hyperspace.Indexes();
Assert.Equal(1, indexes.Count());
Assert.Equal(States.Deleted, indexes.SelectExpr("state").First()[0]);
// Restore the index to active state and verify it is back.
_hyperspace.RestoreIndex(_sampleIndexName);
indexes = _hyperspace.Indexes();
Assert.Equal(1, indexes.Count());
Assert.Equal(States.Active, indexes.SelectExpr("state").First()[0]);
// Delete and vacuum the index, then verify it is gone.
_hyperspace.DeleteIndex(_sampleIndexName);
_hyperspace.VacuumIndex(_sampleIndexName);
Assert.Equal(0, _hyperspace.Indexes().Count());
}
public void TestWord2VecModel()
{
DataFrame documentDataFrame =
_spark.Sql("SELECT split('Hi I heard about Spark', ' ') as text");
Word2Vec word2vec = new Word2Vec()
.SetInputCol("text")
.SetOutputCol("result")
.SetMinCount(1);
Word2VecModel model = word2vec.Fit(documentDataFrame);
const int expectedSynonyms = 2;
DataFrame synonyms = model.FindSynonyms("Hi", expectedSynonyms);
Assert.Equal(expectedSynonyms, synonyms.Count());
synonyms.Show();
using (var tempDirectory = new TemporaryDirectory())
{
string savePath = Path.Join(tempDirectory.Path, "word2vecModel");
model.Save(savePath);
Word2VecModel loadedModel = Word2VecModel.Load(savePath);
Assert.Equal(model.Uid(), loadedModel.Uid());
}
}
private static void ShowDatasetInfo(DataFrame df)
{
//Print first rows and schema
df.Show();
df.PrintSchema();
var numRows = df.Count();
Console.WriteLine($"Found {numRows} movies in movie dataset");
var directorsCount = df.Select("director_name").Distinct().Count();
Console.WriteLine($"Found {directorsCount} diferent Directors in movie dataset");
var maxActorDF = df.GroupBy("actor_1_name").Count().OrderBy(Desc("count"));
maxActorDF.Explain();
maxActorDF.Show();
var actorWithMostAppearances = maxActorDF.First().GetAs <String>(0);
Console.WriteLine($"Actor with most appearances is {actorWithMostAppearances}");
var mostVotedMovieDF = df.Select("movie_title", "movie_facebook_likes").OrderBy(Desc("movie_facebook_likes"));
mostVotedMovieDF.Show();
var mostVotedMovieRow = mostVotedMovieDF.First();
Console.WriteLine($"Most rated movie is {mostVotedMovieRow.GetAs<String>("movie_title")} " +
$"with {mostVotedMovieRow.GetAs<Int32>("movie_facebook_likes")} votes");
}
public void TestEmailSearchTopNReducerBasics()
{
// Read the sample data.
DataFrame df = _spark
.Read()
.Schema("Id STRING, DisplayName STRING, GivenName STRING, Surname STRING, IMAddress STRING, EmailAddress STRING, RelevanceScore DOUBLE, puser STRING, ptenant STRING")
.Json($"{TestEnvironment.ResourceDirectory}neighbors.json");
// Trim the IMAddress column.
Func <Column, Column> trimIMAddress = Udf <string, string>((str) => str.StartsWith("sip:") ? str.Substring(4) : str);
df = df.WithColumn("IMAddress", trimIMAddress(df["IMAddress"]));
// Reduce
df = df.GroupBy("puser", "ptenant").Agg(CollectList("GivenName").Alias("GivenNames"),
CollectList("Surname").Alias("Surnames"),
CollectList("DisplayName").Alias("DisplayNames"),
CollectList("EmailAddress").Alias("EmailAddresses"),
CollectList("RelevanceScore").Alias("RelevanceScores"),
CollectList("IMAddress").Alias("IMAddresses"));
// Format the output.
df = df.Select(df["puser"],
df["ptenant"],
ConcatWs(";", df["GivenNames"]).Alias("GivenNames"),
ConcatWs(";", df["Surnames"]).Alias("Surnames"),
ConcatWs(";", df["DisplayNames"]).Alias("DisplayNames"),
ConcatWs(";", df["EmailAddresses"]).Alias("EmailAddresses"),
ConcatWs(";", df["RelevanceScores"]).Alias("RelevanceScores"),
ConcatWs(";", df["IMAddresses"]).Alias("IMAddresses"));
Assert.Equal(2, df.Count());
foreach (Row row in df.Collect())
{
string puser = row.GetAs <string>("puser");
Assert.Equal("MSFT", row.GetAs <string>("ptenant"));
Assert.Equal("1101.0;900.0;857.0", row.GetAs <string>("RelevanceScores"));
switch (puser)
{
case "ruih":
Assert.Equal("AliceFN;BobFN;CharlieFN", row.GetAs <string>("GivenNames"));
Assert.Equal("AliceLN;BobLN;CharlieLN", row.GetAs <string>("Surnames"));
Assert.Equal("AliceFN AliceLN;BobFN BobLN;CharlieFN CharlieLN", row.GetAs <string>("DisplayNames"));
Assert.Equal("[email protected];[email protected];[email protected]", row.GetAs <string>("EmailAddresses"));
Assert.Equal("[email protected];[email protected];[email protected]", row.GetAs <string>("IMAddresses"));
break;
case "rui":
Assert.Equal("DougFN;ElvaFN;FrankFN", row.GetAs <string>("GivenNames"));
Assert.Equal("DougLN;ElvaLN;FrankLN", row.GetAs <string>("Surnames"));
Assert.Equal("DougFN DougLN;ElvaFN ElvaLN;FrankFN FrankLN", row.GetAs <string>("DisplayNames"));
Assert.Equal("[email protected];[email protected];[email protected]", row.GetAs <string>("EmailAddresses"));
Assert.Equal("[email protected];[email protected];[email protected]", row.GetAs <string>("IMAddresses"));
break;
default:
throw new Exception($"Unexpected age: {puser}.");
}
}
}
/// <inheritdoc cref="Analyzer{S,M}.ComputeStateFrom"/>
public override Option <FrequenciesAndNumRows> ComputeStateFrom(DataFrame dataFrame)
{
long totalCount = dataFrame.Count();
DataFrame dataFrameFiltered = FilterOptional(Where, dataFrame);
DataFrame binnedDataFrame = BinOptional(BinningUdf, dataFrameFiltered);
binnedDataFrame = binnedDataFrame.Select(Col(Column).Cast("string"))
.Na().Fill(NULL_FIELD_REPLACEMENT)
.GroupBy(Column)
.Count()
.WithColumnRenamed("count", AnalyzersExt.COUNT_COL);
return(new Option <FrequenciesAndNumRows>(new FrequenciesAndNumRows(binnedDataFrame, totalCount)));
}
/// <summary>
/// Validate that a range DataFrame contains only the expected values.
/// </summary>
/// <param name="expectedValues"></param>
/// <param name="dataFrame"></param>
private void ValidateRangeDataFrame(IEnumerable <int> expectedValues, DataFrame dataFrame)
{
Assert.Equal(expectedValues.Count(), dataFrame.Count());
var sortedExpectedValues = new List <int>(expectedValues);
sortedExpectedValues.Sort();
var sortedValues = new List <int>(
dataFrame
// We need to select the "id" column, otherwise Collect() won't show the
// updates made to the DeltaTable.
.Select("id")
.Sort("id")
.Collect()
.Select(row => Convert.ToInt32(row.Get("id"))));
Assert.True(sortedValues.SequenceEqual(expectedValues));
}
private static void ReviewsCleanup(DataFrame dataFrame)
{
Console.WriteLine("Ratings Clean-up");
dataFrame = dataFrame
.Filter(
dataFrame["reviewerID"].IsNotNull()
.And(dataFrame["asin"].IsNotNull())
.And(dataFrame["reviewText"].IsNotNull()));
dataFrame = dataFrame
.WithColumnRenamed("reviewerID", "rid")
.WithColumnRenamed("reviewText", "review_text")
.WithColumnRenamed("unixReviewTime", "unix_time");
dataFrame.Cache();
dataFrame.CreateOrReplaceTempView("ElectronicsReviews");
Console.WriteLine($"Reviews Count: {dataFrame.Count()}");
Console.WriteLine("Done");
Console.WriteLine();
}
private static bool ValidateEntity(DataFrame data)
{
var ret = true;
if (data.Schema().Json != _expectedSchema.Json)
{
Console.WriteLine("Expected Schema Does NOT Match");
Console.WriteLine("Actual Schema: " + data.Schema().SimpleString);
Console.WriteLine("Expected Schema: " + _expectedSchema.SimpleString);
ret = false;
}
if (data.Filter(Col("Date").IsNotNull()).Count() == 0)
{
Console.WriteLine("Date Parsing resulted in all NULL's");
ret = false;
}
if (data.Count() == 0)
{
Console.WriteLine("DataFrame is empty");
ret = false;
}
var amountBySuppliers = data.GroupBy(Col("Supplier")).Sum("Amount")
.Filter(Col("Sum(Amount)") < 25000);
if (amountBySuppliers.Count() > 0)
{
Console.WriteLine("Amounts should only ever be over 25k");
amountBySuppliers.Show();
ret = false;
}
return(ret);
}
public void TestSignaturesV2_3_X()
{
Column col = _df["name"];
col = _df["age"];
DataFrame df = _df.ToDF();
df = df.ToDF("name2", "age2");
StructType schema = _df.Schema();
Assert.NotNull(schema);
_df.PrintSchema();
_df.Explain();
_df.Explain(true);
_df.Explain(false);
Assert.Equal(2, _df.Columns().ToArray().Length);
_df.IsLocal();
_df.IsStreaming();
using (var tempDir = new TemporaryDirectory())
{
// The following is required for *CheckPoint().
_spark.SparkContext.SetCheckpointDir(tempDir.Path);
_df.Checkpoint();
_df.Checkpoint(false);
_df.LocalCheckpoint();
_df.LocalCheckpoint(false);
}
_df.WithWatermark("time", "10 minutes");
_df.Show();
_df.Show(10);
_df.Show(10, 10);
_df.Show(10, 10, true);
_df.Join(_df);
_df.Join(_df, "name");
_df.Join(_df, new[] { "name" });
_df.Join(_df, new[] { "name" }, "outer");
_df.Join(_df, _df["age"] == _df["age"]);
_df.Join(_df, _df["age"] == _df["age"], "outer");
_df.CrossJoin(_df);
_df.SortWithinPartitions("age");
_df.SortWithinPartitions("age", "name");
_df.SortWithinPartitions();
_df.SortWithinPartitions(_df["age"]);
_df.SortWithinPartitions(_df["age"], _df["name"]);
_df.Sort("age");
_df.Sort("age", "name");
_df.Sort();
_df.Sort(_df["age"]);
_df.Sort(_df["age"], _df["name"]);
_df.OrderBy("age");
_df.OrderBy("age", "name");
_df.OrderBy();
_df.OrderBy(_df["age"]);
_df.OrderBy(_df["age"], _df["name"]);
_df.Hint("broadcast");
_df.Hint("broadcast", new[] { "hello", "world" });
_df.Col("age");
_df.ColRegex("age");
_df.As("alias");
_df.Alias("alias");
_df.Select("age");
_df.Select("age", "name");
_df.Select();
_df.Select(_df["age"]);
_df.Select(_df["age"], _df["name"]);
_df.SelectExpr();
_df.SelectExpr("age * 2");
_df.SelectExpr("age * 2", "abs(age)");
_df.Filter(_df["age"] > 21);
_df.Filter("age > 21");
_df.Where(_df["age"] > 21);
_df.Where("age > 21");
_df.GroupBy("age");
_df.GroupBy("age", "name");
_df.GroupBy();
_df.GroupBy(_df["age"]);
_df.GroupBy(_df["age"], _df["name"]);
_df.Rollup("age");
_df.Rollup("age", "name");
_df.Rollup();
_df.Rollup(_df["age"]);
_df.Rollup(_df["age"], _df["name"]);
_df.Cube("age");
_df.Cube("age", "name");
_df.Cube();
_df.Cube(_df["age"]);
_df.Cube(_df["age"], _df["name"]);
_df.Agg(Avg(_df["age"]));
_df.Agg(Avg(_df["age"]), Avg(_df["name"]));
_df.Limit(10);
_df.Union(_df);
_df.UnionByName(_df);
_df.Intersect(_df);
_df.Except(_df);
_df.Sample(0.5);
_df.Sample(0.5, true);
_df.Sample(0.5, false, 12345);
_df.RandomSplit(new[] { 0.2, 0.8 });
_df.RandomSplit(new[] { 0.2, 0.8 }, 12345);
_df.WithColumn("age2", _df["age"]);
_df.WithColumnRenamed("age", "age2");
_df.Drop();
_df.Drop("age");
_df.Drop("age", "name");
_df.Drop(_df["age"]);
_df.DropDuplicates();
_df.DropDuplicates("age");
_df.DropDuplicates("age", "name");
_df.Describe();
_df.Describe("age");
_df.Describe("age", "name");
_df.Summary();
_df.Summary("count");
_df.Summary("count", "mean");
_df.Head(2);
_df.Head();
_df.First();
_df.Take(3).ToArray();
_df.Collect().ToArray();
_df.ToLocalIterator().ToArray();
_df.Count();
_df.Repartition(2);
_df.Repartition(2, _df["age"]);
_df.Repartition(_df["age"]);
_df.Repartition();
_df.RepartitionByRange(2, _df["age"]);
_df.RepartitionByRange(_df["age"]);
_df.Coalesce(1);
_df.Distinct();
_df.Persist();
_df.Cache();
_df.Unpersist();
_df.CreateTempView("view");
_df.CreateOrReplaceTempView("view");
_df.CreateGlobalTempView("global_view");
_df.CreateOrReplaceGlobalTempView("global_view");
}
public void TestEmailSearchSuccessActionReducerBasics()
{
// Read the sample data.
DataFrame df = _spark.Read().Json($"{TestEnvironment.ResourceDirectory}search_actions.json");
// Select the required columns.
df = df.Select("ImpressionId", "ConversationId", "EntityType", "FolderIdList", "ReferenceIdList", "ItemIdList", "ItemImmutableIdList");
// Convert columns of concatenated string to array of strings.
Func <Column, Column> toStringArrayUdf = Udf <string, string[]>((str) => str.Split(';'));
df = df.WithColumn("FolderIdList", toStringArrayUdf(df["FolderIdList"]))
.WithColumn("ReferenceIdList", toStringArrayUdf(df["ReferenceIdList"]))
.WithColumn("ItemIdList", toStringArrayUdf(df["ItemIdList"]))
.WithColumn("ItemImmutableIdList", toStringArrayUdf(df["ItemImmutableIdList"]));
// Apply the ArrayZip function to combine the i-th element of each array.
df = df.Select(df["ConversationId"], df["ImpressionId"], df["EntityType"], ArraysZip(df["FolderIdList"], df["ReferenceIdList"], df["ItemIdList"], df["ItemImmutableIdList"]).Alias("ConcatedColumn"));
// Apply the Explode function to split into multiple rows.
df = df.Select(df["ConversationId"], df["ImpressionId"], df["EntityType"], Explode(df["ConcatedColumn"]).Alias("NewColumn"));
// Create multiple columns.
df = df.WithColumn("FolderId", df["NewColumn"].GetField("FolderIdList"))
.WithColumn("ReferenceId", df["NewColumn"].GetField("ReferenceIdList"))
.WithColumn("ItemId", df["NewColumn"].GetField("ItemIdList"))
.WithColumn("ItemImmutableId", df["NewColumn"].GetField("ItemImmutableIdList"))
.Select("ConversationId", "ImpressionId", "EntityType", "FolderId", "ItemId", "ReferenceId", "ItemImmutableId");
// Check the results.
Assert.Equal(3, df.Count());
int i = 0;
foreach (Row row in df.Collect())
{
string impressionId = row.GetAs <string>("ImpressionId");
string conversationId = row.GetAs <string>("ConversationId");
string entityType = row.GetAs <string>("EntityType");
Assert.Equal("Imp1", impressionId);
Assert.Equal("DD8A6B40-B4C9-426F-8194-895E9053077C", conversationId);
Assert.Equal("Message", entityType);
string folderId = row.GetAs <string>("FolderId");
string itemId = row.GetAs <string>("ItemId");
string referenceId = row.GetAs <string>("ReferenceId");
string itemImmutableId = row.GetAs <string>("ItemImmutableId");
if (i == 0)
{
Assert.Equal("F1", folderId);
Assert.Equal("ItemId1", itemId);
Assert.Equal("R1", referenceId);
Assert.Equal("ItemImmutableId1", itemImmutableId);
}
else if (i == 1)
{
Assert.Equal("F2", folderId);
Assert.Equal("ItemId2", itemId);
Assert.Equal("R2", referenceId);
Assert.Equal("ItemImmutableId2", itemImmutableId);
}
else if (i == 2)
{
Assert.Equal("F3", folderId);
Assert.Equal("ItemId3", itemId);
Assert.Equal("R3", referenceId);
Assert.Equal("ItemImmutableId3", itemImmutableId);
}
else
{
throw new Exception(string.Format("Unexpected row: ConversationId={0}, ImpressionId={1}", conversationId, impressionId));
}
i++;
}
}
private static void MetadataCleanup(DataFrame dataFrame)
{
Console.WriteLine("Metadata Clean-up");
var priceCleanup = Udf <string, float>(
p =>
{
if (!string.IsNullOrEmpty(p))
{
var index = 0;
for (var i = 0; i < p.Length; i++)
{
if (char.IsDigit(p[i]))
{
index = i;
break;
}
}
if (float.TryParse(p.Substring(index), out var result))
{
return(result);
}
}
return(-1f);
});
var dateCleanup = Udf <string, double>(
d =>
{
if (!string.IsNullOrEmpty(d) && DateTime.TryParse(d, out var result))
{
return((result.ToUniversalTime() - new DateTime(1970, 1, 1)).TotalSeconds);
}
return(-1L);
});
var rankCleanup = Udf <string, long>(
r =>
{
if (!string.IsNullOrEmpty(r))
{
var regex = new Regex(@"\d+(,\d+)*", RegexOptions.Singleline);
var match = regex.Match(r);
if (match.Success && long.TryParse(match.Value.Replace(",", string.Empty), out var result))
{
return(result);
}
}
return(-1L);
});
dataFrame = dataFrame
.Filter(
dataFrame["asin"].IsNotNull()
.And(dataFrame["title"].IsNotNull())
.And(dataFrame["main_cat"].IsNotNull())
.And(dataFrame["brand"].IsNotNull())
.And(Not(dataFrame["main_cat"].IsIn("Grocery", "Pet Supplies", "Baby", "Books", "Appstore for Android", "Gift Cards"))));
dataFrame = dataFrame
.WithColumn("clean_price", priceCleanup(dataFrame["price"]))
.WithColumn("clean-date", dateCleanup(dataFrame["date"]))
.WithColumn("clean-rank", rankCleanup(dataFrame["rank"]))
.Drop(dataFrame["price"])
.Drop(dataFrame["date"])
.Drop(dataFrame["rank"])
.WithColumnRenamed("clean_price", "price")
.WithColumnRenamed("clean-date", "unixTime")
.WithColumnRenamed("clean-rank", "rank");
dataFrame.Cache();
dataFrame.CreateOrReplaceTempView("ElectronicsMetadata");
Console.WriteLine($"Metadata Count: {dataFrame.Count()}");
Console.WriteLine("Done");
Console.WriteLine();
}
