session, com_stmt: fetch all rows during EXECUTE command

[YangKeao]

What problem does this PR solve?

Issue Number: close #42424, close #41891

What is changed and how it works?

1. Fetch all rows and store them in the fetchedRows in the ResultSet.
2. Close the result set in EXECUTE command, so we don't need to close it again in the following commands (FETCH, RESET and CLOSE)

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No code

Test with the following a little bigger table:

import java.math.BigDecimal;
import java.sql.*;

public class Example {
    public static void main(String[] args) throws SQLException, InterruptedException, Exception {
        Connection conn = DriverManager.getConnection("jdbc:mysql://127.0.0.1:4000/test?useCursorFetch=true&useServerPrepStmts=true&useSSL=false", "root", "");

        conn.setAutoCommit(false); // must set

        conn.prepareStatement("drop table if exists t").execute();
        conn.prepareStatement("create table t (id int primary key, id_2 int)").execute();

        for (int i = 0; i < 10000; i++) {
            PreparedStatement stmt = conn.prepareStatement("insert into t values (?, ?)");
            stmt.setInt(1, i);
            stmt.setInt(2, i);
            stmt.execute();
        }

        // submit a statement with more arguments:
        PreparedStatement statement1 = conn.prepareStatement("" +
                "select * " +
                "from " +
                "t " +
                "where id > ? and id_2 > ?");
        statement1.setFetchSize(500);
        statement1.setBigDecimal(1, new BigDecimal(2));
        statement1.setBigDecimal(2, new BigDecimal(2));
        ResultSet rs_1 = statement1.executeQuery();

        // submit a statement with less arguments:
        PreparedStatement statement2 = conn.prepareStatement("" +
                "select * " +
                "from " +
                "t " +
                "where (id > ?)");
        statement2.setFetchSize(500);
        statement2.setBigDecimal(1, new BigDecimal(1));
        ResultSet rs_2 = statement2.executeQuery();


        // fetch the result from the first argument
        int iter = 3;
        while (rs_1.next()) {
            int result = rs_1.getInt(1);

            if (result != iter++ ) {
                throw new Exception("");
            }
        }
        System.out.println(iter);
        iter = 2;
        while (rs_2.next()) {
            int result = rs_2.getInt(1);

            if (result != iter++ ) {
                throw new Exception("");
            }
        }
        System.out.println(iter);

        conn.commit();
        statement1.close();
        statement2.close();
        conn.close();
    }
}

Side effects

• Performance regression: Consumes more CPU
• Performance regression: Consumes more Memory
• Breaking backward compatibility

It stores all rows of a cursor fetch statement locally, while execuing the EXECUTE command, so it will eat up more memory (related with the size of query).

Release note

Previously, TiDB loads data in `FETCH` command with the `cursorFetch` enabled. Now, TiDB will load all data into the memory in `EXECUTE` command. It'll use more memory according to the size of result. And also, TiDB will never report write conflict in `FETCH` command with read-committed isolation level.

[ti-chi-bot]

[REVIEW NOTIFICATION]

This pull request has been approved by:

• qw4990
• tiancaiamao

To complete the pull request process, please ask the reviewers in the list to review by filling /cc @reviewer in the comment.
After your PR has acquired the required number of LGTMs, you can assign this pull request to the committer in the list by filling /assign @committer in the comment to help you merge this pull request.

The full list of commands accepted by this bot can be found here.

Reviewer can indicate their review by submitting an approval review.
Reviewer can cancel approval by submitting a request changes review.

[YangKeao]

/retest

[zyguan]

cc @jackysp , the other issue of cursor read, we need to tell users that it may cost more memory.

[YangKeao]

The original expected out is wrong. The fetchsize is 0 and the resultSet is not empty, so it shouldn't return 0x82 (which includes mysql.ServerStatusLastRowSend).
I modifed it to the correct answer here: the status is still ServerStatusCursorExists with no answer.

[YangKeao]

/retest

[YangKeao]

rebase the master. @hawkingrei seems to have made some optimization on test speed.

[hawkingrei]

/merge

[ti-chi-bot]

This pull request has been accepted and is ready to merge.

Commit hash: e613832

[YangKeao]

/retest

[YangKeao]

/retest

[ti-chi-bot]

In response to a cherrypick label: new pull request created to branch release-6.5: #42518.

[ti-chi-bot]

In response to a cherrypick label: new pull request created to branch release-7.0: #42519.

[tiancaiamao]

close #38116