Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-6841

test decimal math

Details

    • Task
    • Status: Open (View Workflow)
    • Major
    • Resolution: Unresolved
    • N/A
    • None
    • None

    Description

      Even after all these years there is a steady stream of bugs in decimal math.
      We need a proactive approach instead, to try to find as many decimal bugs as
      possible.

      Suggestion: RQG testing of the expression

      ABS(CAST(a + b AS DOUBLE)/(CAST(a AS DOUBLE) + CAST(b AS DOUBLE)) - 1) < 0.0001

      where RQG randomly generates a and b, and the operation is one of +,
      -, *, /, mod.

      One should try also very small numbers (0.00...{~50 times}...01) as well as
      large ones.

      Attachments

        Issue Links

          relates to

          Bug - A problem which impairs or prevents the functions of the product. MDEV-6851 Decimal multiplication loses precision for no good reason

          • Minor - Minor loss of function, or other problem where easy workaround is present.
          • Open

          Bug - A problem which impairs or prevents the functions of the product. MDEV-6852 Sequence of decimal division and multiplications returns 0 instead of the real result

          • Minor - Minor loss of function, or other problem where easy workaround is present.
          • Open

          Bug - A problem which impairs or prevents the functions of the product. MDEV-6853 Unexpected zero result on 1 % <very small decimal> (wrong default precision of the returned result?)

          • Minor - Minor loss of function, or other problem where easy workaround is present.
          • Open

          Bug - A problem which impairs or prevents the functions of the product. MDEV-6856 <negative value> MOD <fractional divisor> returns negative zero

          • Minor - Minor loss of function, or other problem where easy workaround is present.
          • Open

          Activity

            Ascending order - Click to sort in descending order
            elenst Elena Stepanova added a comment -

            This query works okay in 5.5.40 candidate, but wrong in MySQL 5.5 (apparently that's the same problem as the one serg solved in the post-merge fix). 

            MySQL [test]> select 0.0000000001 % 1;
            		 
            +------------------+
            		 
            | 0.0000000001 % 1 |
            		 
            +------------------+
            		 
            |     0.1000000000 |
            		 
            +------------------+
            		 
            1 row in set (0.00 sec)

            elenst Elena Stepanova added a comment - This query works okay in 5.5.40 candidate, but wrong in MySQL 5.5 (apparently that's the same problem as the one serg solved in the post-merge fix). MySQL [test]> select 0.0000000001 % 1; + ------------------+ | 0.0000000001 % 1 | + ------------------+ | 0.1000000000 | + ------------------+ 1 row in set (0.00 sec)
            serg Sergei Golubchik added a comment -

            Correction

            0.0001 error margin is rather arbitrary. It should depend on operands and on the operation itself. Here, revised queries:

            ABS(CAST(a + b AS DOUBLE)/(CAST(a AS DOUBLE) + CAST(b AS DOUBLE)) - 1) < POW(0.1, GREATEST(LENGTH(REGEXP_SUBSTR(a, '(?<=[.]).*')), LENGTH(REGEXP_SUBSTR(b, '(?<=[.]).*'))))
            		 
            ABS(CAST(a - b AS DOUBLE)/(CAST(a AS DOUBLE) - CAST(b AS DOUBLE)) - 1) < POW(0.1, GREATEST(LENGTH(REGEXP_SUBSTR(a, '(?<=[.]).*')), LENGTH(REGEXP_SUBSTR(b, '(?<=[.]).*'))))
            		 
            ABS(CAST(a * b AS DOUBLE)/(CAST(a AS DOUBLE) * CAST(b AS DOUBLE)) - 1) < POW(0.1, LENGTH(REGEXP_SUBSTR(a, '(?<=[.]).*')) + LENGTH(REGEXP_SUBSTR(b, '(?<=[.]).*')))
            		 
            ABS(CAST(a / b AS DOUBLE)/(CAST(a AS DOUBLE) / CAST(b AS DOUBLE)) - 1) < POW(0.1, LENGTH(REGEXP_SUBSTR(a, '(?<=[.]).*')) + @@div_precision_increment)

            serg Sergei Golubchik added a comment - Correction 0.0001 error margin is rather arbitrary. It should depend on operands and on the operation itself. Here, revised queries: ABS ( CAST (a + b AS DOUBLE )/( CAST (a AS DOUBLE ) + CAST (b AS DOUBLE )) - 1) < POW(0.1, GREATEST(LENGTH(REGEXP_SUBSTR(a, '(?<=[.]).*' )), LENGTH(REGEXP_SUBSTR(b, '(?<=[.]).*' )))) ABS ( CAST (a - b AS DOUBLE )/( CAST (a AS DOUBLE ) - CAST (b AS DOUBLE )) - 1) < POW(0.1, GREATEST(LENGTH(REGEXP_SUBSTR(a, '(?<=[.]).*' )), LENGTH(REGEXP_SUBSTR(b, '(?<=[.]).*' )))) ABS ( CAST (a * b AS DOUBLE )/( CAST (a AS DOUBLE ) * CAST (b AS DOUBLE )) - 1) < POW(0.1, LENGTH(REGEXP_SUBSTR(a, '(?<=[.]).*' )) + LENGTH(REGEXP_SUBSTR(b, '(?<=[.]).*' ))) ABS ( CAST (a / b AS DOUBLE )/( CAST (a AS DOUBLE ) / CAST (b AS DOUBLE )) - 1) < POW(0.1, LENGTH(REGEXP_SUBSTR(a, '(?<=[.]).*' )) + @@div_precision_increment)

            People

              elenst Elena Stepanova
              serg Sergei Golubchik
              Votes:
              1 Vote for this issue
              Watchers:
              3 Start watching this issue

              Dates

                Created:
                Updated:

                Git Integration

                  Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.