Documentation
¶
Overview ¶
Package tclextract pulls literal SQL test cases out of SQLite's TCL test scripts (test/*.test in the SQLite source tree).
It recognizes the dominant test commands, which all take the SQL as their second word:
do_execsql_test <name> {SQL} ?{expected}?
do_catchsql_test <name> {SQL} {expected}
do_eqp_test <name> {SQL} {plan}
plus the older style used by e.g. tokenize.test, but only when the body is exactly one execsql/catchsql call:
do_test <name> {catchsql {SQL}} {expected}
do_test <name> {execsql {SQL}} {expected}
Only the name and the SQL block are used; expected values are re-derived from a real SQLite build by cmd/regenerate-parse, so TCL expectations are ignored. Extraction is deliberately conservative: any case whose SQL block is not a plain TCL brace literal — it contains `$` (variable substitution), `[` (command substitution), or `\` (backslash escapes change brace counting) — is skipped and counted, never guessed at.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
Types ¶
type Case ¶
type Case struct {
Name string // TCL test name, or a generated one if the name was dynamic
SQL string // literal SQL, outer braces removed, normalized to end in "\n"
}
Case is one extracted test case.
type Stats ¶
type Stats struct {
Found int // command occurrences seen
Extracted int
SkippedSubst int // SQL contained $, [ or \
SkippedForm int // SQL argument was not a brace literal / malformed command
GeneratedNames int // dynamic test names replaced with file-local ones
}
Stats reports what extraction skipped, so coverage loss is visible.
Source Files