blob: 44c134fbd154c2c1feca4d33ab075703a4d24cec [file] [log] [blame]
// Copyright (C) 2017 GerritForge Ltd
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package com.gerritforge.analytics
import java.io.{File, FileWriter}
import com.gerritforge.analytics.engine.GerritAnalyticsTransformations._
import com.gerritforge.analytics.model.{GerritEndpointConfig, GerritProject, GerritProjectsRDD, ProjectContributionSource}
import org.apache.spark.sql.Row
import org.json4s.JsonDSL._
import org.json4s._
import org.json4s.jackson.JsonMethods.{compact, render}
import org.scalatest.{FlatSpec, Inside, Matchers}
import scala.io.Source
class GerritAnalyticsTransformationsSpec extends FlatSpec with Matchers
with SparkTestSupport with Inside {
"GerritProjects" should "parse JSON into a GerritProject objects" in {
val projects = GerritProjectsRDD(Source.fromString(
""")]}'
|{
| "All-Projects-name": {
| "id": "All-Projects-id",
| },
| "Test-name": {
| "id": "Test-id",
| }
|}
|""".stripMargin)).collect()
projects should have size 2
projects should contain
allOf(GerritProject("All-Projects-id", "All-Projects-name"), GerritProject("Test-id", "Test-name"))
}
"enrichWithSource" should "enrich project RDD object with its source" in {
val projectRdd = sc.parallelize(Seq(GerritProject("project-id", "project-name")))
implicit val config = GerritEndpointConfig("http://somewhere.com")
val projectWithSource = projectRdd
.enrichWithSource
.collect
projectWithSource should have size 1
inside(projectWithSource.head) {
case ProjectContributionSource(projectName, url) => {
projectName should be("project-name")
url should startWith("http://somewhere.com")
}
}
}
"fetchRawContributors" should "fetch file content from the initial list of project names and file names" in {
val line1 = "foo" -> "bar"
val line2 = "foo1" -> "bar1"
val line3 = "foo2" -> "bar2\u0100" // checks UTF-8 as well
val line3b = "foo3" -> "bar3\u0101"
val projectSource1 = ProjectContributionSource("p1", newSource(line1, line2, line3))
val projectSource2 = ProjectContributionSource("p2", newSource(line3b))
val rawContributors = sc.parallelize(Seq(projectSource1, projectSource2))
.fetchRawContributors
.collect
rawContributors should have size (4)
rawContributors should contain allOf(
("p1","""{"foo":"bar"}"""),
("p1","""{"foo1":"bar1"}"""),
("p1", "{\"foo2\":\"bar2\u0100\"}"),
("p2", "{\"foo3\":\"bar3\u0101\"}")
)
}
"transformCommitterInfo" should "transform a DataFrame with project and json to a workable DF with separated columns" in {
import sql.implicits._
val rdd = sc.parallelize(Seq(
("p1","""{"name":"a","email":"a@mail.com","year":2017,"month":9, "day":11, "hour":23, "num_commits":1, "num_files": 2, "num_distinct_files": 2, "added_lines":1, "deleted_lines":1, "last_commit_date":0, "is_merge": false, "commits":[{ "sha1": "e063a806c33bd524e89a87732bd3f1ad9a77a41e", "date":0,"merge":false, "files": ["file1.txt", "file2.txt"]}] }"""),
("p2","""{"name":"b","email":"b@mail.com","year":2017,"month":9, "day":11, "hour":23, "num_commits":428, "num_files": 2, "num_distinct_files": 3, "added_lines":1, "deleted_lines":1, "last_commit_date":1500000000000, "is_merge": true, "commits":[{"sha1":"e063a806c33bd524e89a87732bd3f1ad9a77a41e", "date":0,"merge":true, "files": ["file3.txt", "file4.txt"] },{"sha1":"e063a806c33bd524e89a87732bd3f1ad9a77a41e", "date":1500000000000,"merge":true, "files": ["file1.txt", "file4.txt"]}]}"""),
// last commit is missing hour,day,month,year to check optionality
("p3","""{"name":"c","email":"c@mail.com","num_commits":12,"num_files": 4, "num_distinct_files": 2, "added_lines":1, "deleted_lines":1, "last_commit_date":1600000000000,"is_merge": true,"commits":[{"sha1":"e063a806c33bd524e89a87732bd3f1ad9a77a41e", "date":0,"merge":true, "files": ["file1.txt", "file2.txt"] },{"sha1":"e063a806c33bd524e89a87732bd3f1ad9a77a41e", "date":1600000000000,"merge":true, "files": ["file1.txt", "file2.txt"]}]}""")
))
val df = rdd.toDF("project", "json")
.transformCommitterInfo
df.count should be(3)
val collected = df.collect
df.schema.fields.map(_.name) should contain inOrder (
"project", "author", "email",
"year", "month", "day", "hour",
"num_files", "num_distinct_files", "added_lines", "deleted_lines",
"num_commits", "last_commit_date",
"is_merge")
collected should contain allOf(
Row("p1", "a", "a@mail.com", 2017, 9, 11, 23, 2, 2, 1, 1, 1, 0, false),
Row("p2", "b", "b@mail.com", 2017, 9, 11, 23, 2, 3, 1, 1, 428, 1500000000000L, true),
Row("p3", "c", "c@mail.com", null, null, null, null, 4, 2, 1, 1, 12, 1600000000000L, true)
)
}
"handleAliases" should "enrich the data with author from the alias DF" in {
import spark.implicits._
val aliasDF = sc.parallelize(Seq(
("aliased_author", "aliased_email@aliased_author.com", "")
)).toDF("author", "email", "organization")
val inputSampleDF = sc.parallelize(Seq(
("author_from_name_a", "non_aliased_email@a_mail.com", "a_mail.com"),
("author_from_name_b", "aliased_email@aliased_author.com", "aliased_author.com")
)).toDF("author", "email", "organization")
val expectedDF = sc.parallelize(Seq(
("author_from_name_a", "non_aliased_email@a_mail.com", "a_mail.com"),
("aliased_author", "aliased_email@aliased_author.com", "aliased_author.com")
)).toDF("author", "email", "organization")
val df = inputSampleDF.handleAliases(Some(aliasDF))
df.schema.fields.map(_.name) should contain allOf(
"author", "email", "organization")
df.collect should contain theSameElementsAs expectedDF.collect
}
it should "enrich the data with organization from the alias DF when available" in {
import spark.implicits._
val aliasDF = sc.parallelize(Seq(
("aliased_author_with_organization", "aliased_email@aliased_organization.com", "aliased_organization"),
("aliased_author_empty_organization", "aliased_email@emtpy_organization.com", ""),
("aliased_author_null_organization", "aliased_email@null_organization.com", null)
)).toDF("author", "email", "organization")
val inputSampleDF = sc.parallelize(Seq(
("author_from_name_a", "aliased_email@aliased_organization.com", "aliased_organization.com"),
("author_from_name_b", "aliased_email@emtpy_organization.com", "emtpy_organization.com"),
("author_from_name_c", "aliased_email@null_organization.com", "null_organization.com")
)).toDF("author", "email", "organization")
val expectedDF = sc.parallelize(Seq(
("aliased_author_with_organization", "aliased_email@aliased_organization.com", "aliased_organization"),
("aliased_author_empty_organization", "aliased_email@emtpy_organization.com", "emtpy_organization.com"),
("aliased_author_null_organization", "aliased_email@null_organization.com", "null_organization.com")
)).toDF("author", "email", "organization")
val df = inputSampleDF.handleAliases(Some(aliasDF))
df.schema.fields.map(_.name) should contain allOf(
"author", "email", "organization")
df.collect should contain theSameElementsAs expectedDF.collect
}
it should "return correct columns when alias DF is defined" in {
import spark.implicits._
val inputSampleDF = sc.parallelize(Seq(
("author_name", "email@mail.com", "an_organization")
)).toDF("author", "email", "organization")
val aliasDF = sc.parallelize(Seq(
("a_random_author", "a_random_email@mail.com", "a_random_organization")
)).toDF("author", "email", "organization")
val df = inputSampleDF.handleAliases(Some(aliasDF))
df.schema.fields.map(_.name) should contain allOf("author", "email", "organization")
}
it should "return correct columns when alias DF is not defined" in {
import spark.implicits._
val inputSampleDF = sc.parallelize(Seq(
("author_name", "email@mail.com", "an_organization")
)).toDF("author", "email", "organization")
val df = inputSampleDF.handleAliases(None)
df.schema.fields.map(_.name) should contain allOf("author", "email", "organization")
}
it should "lowercase aliased organizations" in {
import spark.implicits._
val inputSampleDF = sc.parallelize(Seq(
("author_name", "email@mail.com", "an_organization")
)).toDF("author", "email", "organization")
val aliasDF = sc.parallelize(Seq(
("author_name", "email@mail.com", "OrGaNiZaTiOnToBeLoWeRcAsEd")
)).toDF("author", "email", "organization")
val df = inputSampleDF.handleAliases(Some(aliasDF))
val expectedDF = sc.parallelize(Seq(
("author_name", "email@mail.com", "organizationtobelowercased")
)).toDF("author", "email", "organization")
df.collect should contain theSameElementsAs expectedDF.collect
}
"addOrganization" should "compute organization column from the email" in {
import sql.implicits._
val df = sc.parallelize(Seq(
"",
"@", // corner case
"not an email",
"email@domain-simple",
"email@domain-com.com",
"email@domain-couk.co.uk",
"email@domain-info.info",
"email@mail.companyname-couk.co.uk",
"email@mail.companyname-com.com",
"email@mail.companyname-info.info"
)).toDF("email")
val transformed = df.addOrganization()
transformed.schema.fields.map(_.name) should contain allOf("email", "organization")
transformed.collect should contain allOf(
Row("", ""),
Row("@", ""),
Row("not an email", ""),
Row("email@domain-simple", "domain-simple"),
Row("email@domain-com.com", "domain-com"),
Row("email@domain-couk.co.uk", "domain-couk"),
Row("email@domain-info.info", "domain-info"),
Row("email@mail.companyname-couk.co.uk", "mail.companyname-couk"),
Row("email@mail.companyname-com.com", "mail.companyname-com"),
Row("email@mail.companyname-info.info", "mail.companyname-info")
)
}
"convertDates" should "process specific column from Long to ISO date" in {
// some notable dates converted in UnixMillisecs and ISO format
val DATES = Map(
0L -> "1970-01-01T00:00:00Z",
1500000000000L -> "2017-07-14T02:40:00Z",
1600000000000L -> "2020-09-13T12:26:40Z")
import sql.implicits._
val df = sc.parallelize(Seq(
("a", 0L, 1),
("b", 1500000000000L, 2),
("c", 1600000000000L, 3))).toDF("name", "date", "num")
val converted = df
.convertDates("date")
converted.collect should contain allOf(
Row("a", DATES(0), 1),
Row("b", DATES(1500000000000L), 2),
Row("c", DATES(1600000000000L), 3)
)
}
private def newSource(contributorsJson: JObject*): String = {
val tmpFile = File.createTempFile(System.getProperty("java.io.tmpdir"),
s"${getClass.getName}-${System.nanoTime()}")
val out = new FileWriter(tmpFile)
contributorsJson.foreach(json => out.write(compact(render(json)) + '\n'))
out.close
tmpFile.toURI.toString
}
}