summaryrefslogtreecommitdiff
path: root/test/fts1k.test
diff options
context:
space:
mode:
authorHans-Christoph Steiner <hans@eds.org>2012-03-30 20:42:12 -0400
committerHans-Christoph Steiner <hans@eds.org>2012-03-30 20:42:12 -0400
commit7bb481fda9ecb134804b49c2ce77ca28f7eea583 (patch)
tree31b520b9914d3e2453968abe375f2c102772c3dc /test/fts1k.test
Imported Upstream version 2.0.3
Diffstat (limited to 'test/fts1k.test')
-rw-r--r--test/fts1k.test69
1 files changed, 69 insertions, 0 deletions
diff --git a/test/fts1k.test b/test/fts1k.test
new file mode 100644
index 0000000..35b94d2
--- /dev/null
+++ b/test/fts1k.test
@@ -0,0 +1,69 @@
+# 2007 March 28
+#
+# The author disclaims copyright to this source code.
+#
+#*************************************************************************
+# This file implements regression tests for SQLite library. The focus
+# of this script is testing isspace/isalnum/tolower problems with the
+# FTS1 module. Unfortunately, this code isn't a really principled set
+# of tests, because it is impossible to know where new uses of these
+# functions might appear.
+#
+# $Id: fts1k.test,v 1.2 2007/12/13 21:54:11 drh Exp $
+#
+
+set testdir [file dirname $argv0]
+source $testdir/tester.tcl
+
+# If SQLITE_ENABLE_FTS1 is defined, omit this file.
+ifcapable !fts1 {
+ finish_test
+ return
+}
+
+# Tests that startsWith() (calls isspace, tolower, isalnum) can handle
+# hi-bit chars. parseSpec() also calls isalnum here.
+do_test fts1k-1.1 {
+ execsql "CREATE VIRTUAL TABLE t1 USING fts1(content, \x80)"
+} {}
+
+# Additionally tests isspace() call in getToken(), and isalnum() call
+# in tokenListToIdList().
+do_test fts1k-1.2 {
+ catch {
+ execsql "CREATE VIRTUAL TABLE t2 USING fts1(content, tokenize \x80)"
+ }
+ sqlite3_errmsg $DB
+} "unknown tokenizer: \x80"
+
+# Additionally test final isalnum() in startsWith().
+do_test fts1k-1.3 {
+ execsql "CREATE VIRTUAL TABLE t3 USING fts1(content, tokenize\x80)"
+} {}
+
+# The snippet-generation code has calls to isspace() which are sort of
+# hard to get to. It finds convenient breakpoints by starting ~40
+# chars before and after the matched term, and scanning ~10 chars
+# around that position for isspace() characters. The long word with
+# embedded hi-bit chars causes one of these isspace() calls to be
+# exercised. The version with a couple extra spaces should cause the
+# other isspace() call to be exercised. [Both cases have been tested
+# in the debugger, but I'm hoping to continue to catch it if simple
+# constant changes change things slightly.
+#
+# The trailing and leading hi-bit chars help with code which tests for
+# isspace() to coalesce multiple spaces.
+
+set word "\x80xxxxx\x80xxxxx\x80xxxxx\x80xxxxx\x80xxxxx\x80xxxxx\x80"
+set phrase1 "$word $word $word target $word $word $word"
+set phrase2 "$word $word $word target $word $word $word"
+
+db eval {CREATE VIRTUAL TABLE t4 USING fts1(content)}
+db eval "INSERT INTO t4 (content) VALUES ('$phrase1')"
+db eval "INSERT INTO t4 (content) VALUES ('$phrase2')"
+
+do_test fts1k-1.4 {
+ execsql {SELECT rowid, length(snippet(t4)) FROM t4 WHERE t4 MATCH 'target'}
+} {1 111 2 117}
+
+finish_test