Bug 7962 – std.regex: Captures.length() returns incorrect value

It seems that Captures.length() from std.regex returns bogus lengths for empty captures (I might be using it improperly). Here's a simple example, the second match object claims length 2 while being empty. sh-4.2$ uname Linux sh-4.2$ dmd | grep "DMD" DMD64 D Compiler v2.059 sh-4.2$ cat test.d import std.regex; import std.stdio; int main() { auto r = regex(r"(@\w+)", "g"); writeln("match first string -- looks ok"); string text1 = "@This is a test"; auto m1 = match(text1, r); auto myCapts1 = m1.captures; writeln("content: ", myCapts1); writeln("is empty? ", myCapts1.empty()); writeln("length: ", myCapts1.length()); writeln(""); writeln("match second string -- length seems wrong"); string text2 = "(his) is a test"; auto m2 = match(text2, r); auto myCapts2 = m2.captures; writeln("content: ", myCapts2); writeln("is empty? ", myCapts2.empty()); writeln("length : ", myCapts2.length()); return 0; } sh-4.2$ dmd test.d sh-4.2$ ./test match first string -- looks ok content: ["@This", "@This"] is empty? false length: 2 match second string -- length seems wrong content: [] is empty? true length : 2

(In reply to comment #0) > It seems that Captures.length() from std.regex returns bogus lengths for empty > captures (I might be using it improperly). Here's a simple example, the second > match object claims length 2 while being empty. > Thanks for the detailed info. The problem is that Captures.length is hardwired to the number of sub captures + 1 which is fine as long as there is a match. For now you may use m.empty ? 0 : m.length as a workaround. Pull: https://github.com/D-Programming-Language/phobos/pull/548 P.S. Usage of e.g. length() is deprecated and fails with -property switch. Omit parens for @property members.

(In reply to comment #1) > > Thanks for the detailed info. The problem is that Captures.length is hardwired > to the number of sub captures + 1 which is fine as long as there is a match. > > For now you may use m.empty ? 0 : m.length as a workaround. Thanks a lot for your quick response and I will use your suggested workaround. > P.S. Usage of e.g. length() is deprecated and fails with -property switch. Omit > parens for @property members. Thanks for pointing this out and also the -property switch. Sorry, to be off-topic (with regard to this bug) but incidentally, it doesn't seem to do anything sh-4.2$ fgrep 'length()' test.d writeln("length: ", myCapts1.length()); writeln("length : ", myCapts2.length()); sh-4.2$ dmd -property test.d sh-4.2$ Does it need to be combined with another switch to "enforce"?

Commits pushed to master at https://github.com/D-Programming-Language/phobos https://github.com/D-Programming-Language/phobos/commit/6c9414c8c9f1df9f2c42229cf1b7eb8bd90d076b fix Issue 7962 Captures.length() returns incorrect value https://github.com/D-Programming-Language/phobos/commit/70ad8e4aaeda91d8c1db19b07ac3e3c3db2d23a6 Merge pull request #548 from blackwhale/fix-captures fix Issue 7962 Captures.length() returns incorrect value

>Does it need to be combined with another switch to "enforce"? I think it's -w switch that stands for warnings as errors. Or -wi to just enable them. Strange as I thought -property was independent entity.