INNER JOIN differences with h2 database and postgresql database - database

Question

So I've changed my codes database from h2 to postgresql and I've noticed that the Inner Join call that is used in h2 is not giving the same results when I call it in postgresql. I've done research and after testing, I found out that the left join and other joins work perfectly it's only inner join giving me a different result. So, to get both output csv files to match, would I have to change the whole structure of the table or is there something similiar in postgresql that i'm overlooking?
public void doAllWork(int type, Connection conn, Statement st) {
try {
if (type == 1) {
st.execute("DROP TABLE IF EXISTS COMBINEDDATA;"); //USING DISTINCT TO EXCLUDE DUPLICATE RECORDS
st.execute("ANALYZE");
st.execute("CREATE TABLE COMBINEDDATA AS \n"
+ "SELECT DISTINCT E.DATA1, E.DATA2, E.DATA3, E.DATA4, E.DATA5, E.DATA6, \n"
+ "E.DATA7, E.DATA8, E.DATA9, E.DATA10, E.DATA11, E.DATA12, E.DATA13, E.DATA14, E.DATA15, E.DATA16, E.DATA17, \n"
+ "E.DATA18, E.DATA19, E.DATA21, E.DATA26, E.DATA27, E.DATA28, E.DATA29, \n"
+ "E.DATA30, E.DATA31, E.DATA32, E.DATA34, E.DATA35, E.DATA36, E.DATA37, E.DATA38, \n"
+ " C.CHAIN20, C.CHAIN33, C.CHAIN22, \n "
+ "D.DAT2, D.DAT3, D.DAT4, D.DAT7, D.DAT11, D.DAT9, D.DAT5, \n "
+ "E.DATA39, E.DATA40, E.DATA41 FROM rawData AS E \n"
+ "RIGHT JOIN CHAINDATA AS C \n"
+ "ON E.DATA7 = c.CHAIN2\n"
+ "AND E.DATA11 = c.CHAIN4\n"
+ "AND E.DATA21 = c.CHAIN10\n"
+ "AND E.DATA22 = c.CHAIN11\n"
+ "RIGHT JOIN DATDATA AS D\n"
+ "ON E.DATA7 = D.DAT18\n"
+ "AND E.DATA11 = D.DAT21\n"
+ "AND UCASE(E.DATA6) = UCASE(D.DAT17)\n"
+ "AND UCASE(E.DATA10) = UCASE(D.DAT20)\n"
+ "AND UCASE(E.DATA5) = UCASE(D.DAT16)\n"
+ "AND UCASE(E.DATA9) = UCASE(D.DAT19)\n"
+ "AND E.DATA20 = D.DAT22");
} else if (type == 2) {
st.execute("DROP TABLE IF EXISTS COMBINEDDATA2;");
st.execute("ANALYZE");
st.execute("CREATE TABLE COMBINEDDATA2 AS \n"
+ "SELECT DISTINCT E.DATA1, E.DATA2, E.DATA3, E.DATA4, E.DATA5, E.DATA6, \n"
+ "E.DATA7, E.DATA8, E.DATA9, E.DATA10, E.DATA11, E.DATA12, E.DATA13, E.DATA14, E.DATA15, E.DATA16, E.DATA17, \n"
+ "E.DATA18, E.DATA19, E.DATA21, E.DATA26, E.DATA27, E.DATA28, E.DATA29, \n"
+ "E.DATA30, E.DATA31, E.DATA32, E.DATA34, E.DATA35, E.DATA36, E.DATA37, E.DATA38, \n"
+ " C.CHAIN20, C.CHAIN33, C.CHAIN22, \n "
+ "D.DAT2, D.DAT3, D.DAT4, D.DAT7, D.DAT11, D.DAT9, D.DAT5, \n "
+ "E.DATA39, E.DATA40, E.DATA41 FROM rawData AS E \n"
+ "LEFT JOIN CHAINDATA AS C \n"
+ "ON E.DATA7 = c.CHAIN2\n"
+ "AND E.DATA11 = c.CHAIN4\n"
+ "AND E.DATA21 = c.CHAIN10\n"
+ "AND E.DATA22 = c.CHAIN11\n"
+ "LEFT JOIN DATDATA AS D\n"
+ "ON E.DATA7 = D.DAT18\n"
+ "AND E.DATA11 = D.DAT21\n"
+ "AND UCASE(E.DATA6) = UCASE(D.DAT17)\n"
+ "AND UCASE(E.DATA10) = UCASE(D.DAT20)\n"
+ "AND UCASE(E.DATA5) = UCASE(D.DAT16)\n"
+ "AND UCASE(E.DATA9) = UCASE(D.DAT19)\n"
+ "AND E.DATA20 = D.DAT22");
}
System.out.println("here");
if (type == 1) {
String dir = System.getProperty("user.dir");
st.executeUpdate("CALL CSVWRITE('" + dir + "\\OnlyMatching.csv', 'SELECT * FROM COMBINEDDATA','charset=UTF-8');"); //,
} else if (type == 2) {
String dir = System.getProperty("user.dir");
st.executeUpdate("CALL CSVWRITE('" + dir + "\\AllNonMatching.csv', 'SELECT * FROM COMBINEDDATA2','charset=UTF-8');");
}
} catch (Exception ex) {
Logger.getLogger(RyderCombinerGUI.class.getName()).log(Level.SEVERE, null, ex);
}
}
In the above snippet, the second loop with the left join works the same on h2 and postgresql, but the inner join loop returns something different.
Ex)
This is the output csv file using the h2 database.
And this is the output using postgresql database
Thanks in advance.

Answer 1

Assuming that you run the same ANSI-compliant query, with the same underlying data, in both H2 and Postgres, you should get the same result. There is nothing whatosever different about the behavior of INNER JOIN in either database.
But a quick search for ORDER BY in your code dump revealed that you are not doing any ordering in your queries. I noticed that Postgres coincidentally appears to be sorting on the data1 column, while H2 does not appear to be sorting at all. I suggest that the result sets are identical from the point of view of unsorted sets.
In general, if you expect a cetain ordering in your result set, you need to use ORDER BY in the query which generates that data. So if you add ORDER BY data1 to both queries, I expect the results will appear the same for both H2 and Postgres.