Attached is a patch that changes the "find word" test in string_cs.t
to use the find_cclass and find_not_cclass opcodes to locate word
boundaries.
This test doesn't work exactly the way Leo has envisioned it (creating
a special .CCLASS_* value for word boundaries), but it demonstrates
how I think PIR programmers will typically locate word boundaries.
Sent as a patch rather than applied directly to give others an
opportunity to review and comment.
Pm
Index: t/op/string_cs.t
===================================================================
--- t/op/string_cs.t (revision 10026)
+++ t/op/string_cs.t (working copy)
@@ -244,25 +244,26 @@
0 2 5 7 ok
OUTPUT
-SKIP: {
- skip('TODO wordboundary has no cclass', 1);
output_is( <<'CODE', <<OUTPUT, "find a word_boundary");
+.include "cclass.pasm"
set S0, "_ab 09z"
set I0, 0
+ length I1, S0
lp:
- find_cclass I0, .CCLASS_???, S0, I0, I1
+ find_cclass I0, .CCLASS_WORD, S0, I0, I1
print I0
print " "
- eq I0, -1, done
- inc I0
- branch lp
+ eq I0, I1, done
+ find_not_cclass I0, .CCLASS_WORD, S0, I0, I1
+ print I0
+ print " "
+ ne I0, I1, lp
done:
print "ok\n"
end
CODE
-0 2 3 6 7 ok
+0 3 4 7 ok
OUTPUT
-}
output_is( <<'CODE', <<OUTPUT, "trans_charset_s_s_i");
set S0, "abc"
