$ re2c -c -o example.c -Wcondition-order wcondition_order.re
$
- $ gcc -o example example.c
+ $ g++ -o example example.c
$ ./example
aaaa,bbb!
$
- $ gcc -o example -DREVERSED_CONDITION_ORDER example.c
+ $ g++ -o example -DREVERSED_CONDITION_ORDER example.c
$ ./example
aaaa,bbb!
re2c: warning: line 31: looks like you use hardcoded numbers instead of autogenerated condition names: better add
'/*!types:re2c*/' directive or '-t, --type-header' option and don't rely on fixed condition order. [-Wcondition-order]
$
- $ gcc -o example example.c
+ $ g++ -o example example.c
$ ./example
aaaa,bbb!
$
- $ gcc -o example -DREVERSED_CONDITION_ORDER example.c
+ $ g++ -o example -DREVERSED_CONDITION_ORDER example.c
$ ./example
error
--- /dev/null
+False alarms
+~~~~~~~~~~~~
+
+In many cases matching empty string makes perfect sense:
+
+* It might be used as a non-consuming default rule.
+
+* It might be used to lex an optional lexeme: if lexeme rules didn't match,
+ lexer must jump to another block and resume lexing at the same input position.
+
+Or any other useful examples you can invent.
+All these cases are perfectly sane.
+If ``[-Wmatch-empty-string]`` becomes annoying, use ``[-Wno-match-empty-string]``.
+Maybe re2c should move this warning to some paranoid level.
+
+
--- /dev/null
+Real-world examples
+~~~~~~~~~~~~~~~~~~~
+
+In general, it is a common mistake to use ``*`` instead of ``+`` in repetitions.
+That is, to accept zero or more repetitions instead of one or more.
+
+Typos in definitions
+....................
+
+Here is the skeleton of REXX lexer (the very lexer which motivated Peter to write re2c ``:)``).
+
+.. code-block:: cpp
+ :number-lines:
+
+ /*!re2c
+ all = [\000-\377];
+ eof = [\000];
+ any = all\eof;
+ letter = [a-z]|[A-Z];
+ digit = [0-9];
+ symchr = letter|digit|[.!?_];
+ const = (digit|[.])symchr*([eE][+-]?digit+)?;
+ simple = (symchr\(digit|[.]))(symchr\[.])*;
+ stem = simple [.];
+ symbol = symchr*;
+ sqstr = ['] ((any\['\n])|(['][']))* ['];
+ dqstr = ["] ((any\["\n])|(["]["]))* ["];
+ str = sqstr|dqstr;
+ ob = [ \t]*;
+ not = [\\~];
+ A = [aA];
+ B = [bB];
+ C = [cC];
+ D = [dD];
+ E = [eE];
+ F = [fF];
+ G = [gG];
+ H = [hH];
+ I = [iI];
+ J = [jJ];
+ K = [kK];
+ L = [lL];
+ M = [mM];
+ N = [nN];
+ O = [oO];
+ P = [pP];
+ Q = [qQ];
+ R = [rR];
+ S = [sS];
+ T = [tT];
+ U = [uU];
+ V = [vV];
+ W = [wW];
+ X = [xX];
+ Y = [yY];
+ Z = [zZ];
+
+ "\n" {}
+ "|" ob "|" {}
+ "+" {}
+ "-" {}
+ "*" {}
+ "/" {}
+ "%" {}
+ "/" ob "/" {}
+ "*" ob "*" {}
+ "=" {}
+ not ob "=" | "<" ob ">" | ">" ob "<" {}
+ ">" {}
+ "<" {}
+ ">" ob "=" | not ob "<" {}
+ "<" ob "=" | not ob ">" {}
+ "=" ob "=" {}
+ not ob "=" ob "=" {}
+ ">" ob ">" {}
+ "<" ob "<" {}
+ ">" ob ">" ob "=" | not ob "<" ob "<" {}
+ "<" ob "<" ob "=" | not ob ">" ob ">" {}
+ "&" {}
+ "|" {}
+ "&" ob "&" {}
+ not {}
+ ":" {}
+ "," {}
+ "(" {}
+ ")" {}
+ ";" {}
+ A D D R E S S {}
+ A R G {}
+ C A L L {}
+ D O {}
+ D R O P {}
+ E L S E {}
+ E N D {}
+ E X I T {}
+ I F {}
+ I N T E R P R E T {}
+ I T E R A T E {}
+ L E A V E {}
+ N O P {}
+ N U M E R I C {}
+ O P T I O N S {}
+ O T H E R W I S E {}
+ P A R S E {}
+ P R O C E D U R E {}
+ P U L L {}
+ P U S H {}
+ Q U E U E {}
+ R E T U R N {}
+ S A Y {}
+ S E L E C T {}
+ S I G N A L {}
+ T H E N {}
+ T R A C E {}
+ W H E N {}
+ O F F {}
+ O N {}
+ B Y {}
+ D I G I T S {}
+ E N G I N E E R I N G {}
+ E R R O R {}
+ E X P O S E {}
+ F A I L U R E {}
+ F O R {}
+ F O R E V E R {}
+ F O R M {}
+ F U Z Z {}
+ H A L T {}
+ L I N E I N {}
+ N A M E {}
+ N O T R E A D Y {}
+ N O V A L U E {}
+ S C I E N T I F I C {}
+ S O U R C E {}
+ S Y N T A X {}
+ T O {}
+ U N T I L {}
+ U P P E R {}
+ V A L U E {}
+ V A R {}
+ V E R S I O N {}
+ W H I L E {}
+ W I T H {}
+ const {}
+ simple {}
+ stem {}
+ symbol {}
+ str {}
+ str [bB] / (all\symchr) {}
+ str [xX] / (all\symchr) {}
+ eof {}
+ any {}
+ */
+
+```re2c -Wmatch-empty-string``` warns:
+
+.. code-block::
+
+ re2c: warning: line 133: rule matches empty string [-Wmatch-empty-string]
+
+The faulty rule is ``symbol``.
+It is defined as ``symchr*`` and clearly is nullable.
+In this particular example (assuming ASCII encoding) empty match is shadowed by other rules:
+together ``eof`` and ``any`` cover all possible code units.
+So in this case there is no chance of hitting eternal loop.
+
+However, by no means ``symbol`` should be nullable: it makes no sense.
+Sure, it's just a typo and the author meant ``symchr+``.
+
+
+Skipping uninteresting stuff
+............................
+
+Like spaces.
+One often needs to skip variable number of them:
+
+.. code-block:: cpp
+ :number-lines:
+
+ /*!re2c
+ TABS_AND_SPACES = [ \t]*;
+ */
+
+This definition is ok when used inside of another (non-nullable) rule:
+
+.. code-block:: cpp
+ :number-lines:
+
+ /*!re2c
+ TABS_AND_SPACES = [ \t]*;
+ "(" TABS_AND_SPACES ("int" | "integer") TABS_AND_SPACES ")" {}
+ */
+
+However, as a standalone rule it may cause eternal loop on ill-formed input.
+And it's very common to reuse one rule for multiple purposes.
+
+
+
--- /dev/null
+A simple example
+~~~~~~~~~~~~~~~~
+
+``[-Wmatch-empty-string]`` warns when a rule is nullable (matches empty string).
+It was intended to prevent hitting eternal loop in cases like this:
+
+`[wmatch_empty_string.re] <wmatch_empty_string.re>`_
+
+.. include:: wmatch_empty_string.re
+ :code: cpp
+ :number-lines:
+
+The program loops over its arguments (the outer ``for`` loop)
+and tries to lex each argument (the inner ``for`` loop).
+Lexer stops when all input has been consumed and it sees the terminating ``NULL``.
+Arguments must consist of lowercase letters only.
+Generate, compile and run:
+
+.. code-block::
+
+ $ re2c -o example.c -Wmatch-empty-string wmatch_empty_string.re
+ re2c: warning: line 11: rule matches empty string [-Wmatch-empty-string]
+ $ g++ -o example example.c
+ $
+ $ ./example only lowercase letters are allowed
+ argv[1]: only
+ argv[2]: lowercase
+ argv[3]: letters
+ argv[4]: are
+ argv[5]: allowed
+ $
+ $ ./example oh really?
+ argv[1]: oh
+ ^C
+
+The program hangs forever if one of the arguments is ill-formed.
+
+Note that `[-Wundefined-control-flow] <../undefined_control_flow/wundefined_control_flow.html>`_
+has no complaints about this particular case: all input patterns are covered by rules.
+Yet if we add default rule ``*``, lexer won't hang anymore: it will match default rule
+instead of nullable rule.
+
+The fix is easy: make the rule non-nullable (say, ``[a-z]+``) and add default rule ``*``.
+
+
--- /dev/null
+#include <stdio.h>
+
+int main(int argc, char **argv)
+{
+ for (int i = 1; i < argc; ++i) {
+ for (char *YYCURSOR = argv[i];;) {
+ /*!re2c
+ re2c:define:YYCTYPE = char;
+ re2c:yyfill:enable = 0;
+ "\x00" { break; }
+ [a-z]* { continue; }
+ */
+ }
+ printf("argv[%d]: %s\n", i, argv[i]);
+ }
+ return 0;
+}
--- /dev/null
+[-Wmatch-empty-string]
+--------------------------
+
+.. include:: ../home.rst
+.. include:: ../../../contents.rst
+
+.. include:: simple_example.rst
+.. include:: false_alarms.rst
+.. include:: real_world.rst
+
* `[-Wuseless-escape] <useless_escape/wuseless_escape.html>`_
* `[-Wswapped-range] <swapped_range/wswapped_range.html>`_
* `[-Wempty-character-class] <empty_character_class/wempty_character_class.html>`_
-* `[-Wmatch-empty-string] <wmatch_empty_string.html>`_
+* `[-Wmatch-empty-string] <match_empty_string/wmatch_empty_string.html>`_
.. include:: warnings_general.rst
+++ /dev/null
-[-Wmatch-empty-string]
---------------------------
-
-.. include:: home.rst
-
projects / re2c / commitdiff