1. Make sure JDK 8 and Maven 3 are installed
2. git clone https://github.com/guiguan/table-join.git
3. cd table-join

mvn clean install

mvn clean install -DskipTests

java -jar target/*.jar

java -jar target/*.jar t1JsonPath t2JsonPath outputPath

1. Values of z are unique in t1, but not in t2
2. t2 has about 3 times as many rows as t1
3. Each row of json input files is a small JSON object
4. Precision of double is defined in DoubleAdapter.precision, which is set to 6, so 46.1999999 is considered equivalent to 46.2, but not 46.199999. Double is used in this project instead of float because Float.parseFloat can only keep up to 5 decimal places for a floating number. Double gives more freedom on the value of DoubleAdapter.precision

1. GSON is chosen for JSON parsing according to this benchmark for large number of small JSON objects
2. t1 is processed line by line first and a hashtable is constructed for t1.z, which is then used to fast check join criteria of t2 line by line.
3. When processing each line of t2, zz is ignored, and sum aggregation is updated on the fly and no info of t2 is stored otherwise.
4. Finally, results from 3 is sorted according to SUM(t2.y)