fix: parsing DDL with TAB between column name and the column type

[ljwun]

• Do only one thing
• Non breaking API changes
• Tested

What did this pull request do?

Remove tab character (\t) from sqliteSeparator.

sqlite/ddlmod.go

Lines 58 to 79 in 7818dfd

	if sc := string(c); separatorRegexp.MatchString(sc) {
	if c == next {
	buf += sc // Skip escaped quote
	idx++
	} else if quote > 0 {
	quote = 0
	} else {
	quote = c
	}
	} else if quote == 0 {
	if c == '(' {
	bracketLevel++
	} else if c == ')' {
	bracketLevel--
	} else if bracketLevel == 0 {
	if c == ',' {
	result.fields = append(result.fields, strings.TrimSpace(buf))
	buf = ""
	continue
	}
	}
	}

When a tab character (\t) is present between the identifier and its type, separatorRegexp.MatchString(sc) can match and the code toggles the quote state incorrectly. The incorrect quote state causes the parser to ignore commas and therefore fail to split fields properly.

User Case Description

When parsing DDL that uses a tab character (\t) as whitespace between an identifier and its type, the parser can misinterpret the column list and return the wrong number of columns.

Reproduction DDL:

CREATE TABLE `test-e` (`field1`\ttext NOT NULL,`field2`\tinteger NOT NULL)

Expected result should have two columns:

• field1
• field2

Only one column is actually returned (the parser fails to split the column definitions correctly):

• field1

[Shion1305]

LGTM!