Handle NULL values in hash calculation for accurate data comparison

[bienvenuushindi]

Context:

While working with DBDiff to compare data and schema, an issue was identified in the handling of NULL values during data comparison. Specifically, when columns containing NULL values were included in the MD5 hash computation, the results were inconsistent. This issue was due to the behavior of CONCAT in SQL, which returns NULL if any of the values being concatenated are NULL.

Root Cause:

When constructing MD5 hashes for row comparisons, the query builder used CONCAT without handling NULL values. This led to NULL hashes, causing incorrect identification of rows as identical even when there were clear differences.

For example:

SELECT CONCAT('value', NULL);  -- Result is NULL as shown below
+-----------------------+
| CONCAT('value', NULL) |
+-----------------------+
| NULL                  |
+-----------------------+

[kroky]

Original issue found here: https://gitlab.com/tikiwiki/tiki/-/merge_requests/6532#note_2241090253