Fix the plural equation. We changed the formula to a new one years ago!

Source: http://localization-guide.readthedocs.org/en/latest/l10n/pluralforms.html

Hi,

we have included tinygettext in supertuxkart, and have applied a few minor patches over time. A full list can be found on our github: https://github.com/supertuxkart/stk-code/commits/master/src/tinygettext

But here in short the somewhat important and portable ones: Fix missing include for some newer compilers:

--- a/src/tinygettext/language.cpp
+++ b/src/tinygettext/language.cpp
@@ -20,6 +20,7 @@
 #include <map>
 #include <assert.h>
 #include <vector>
+#include <algorithm>

 namespace tinygettext {

Change pt-BR name to be more accurate:

--- a/src/tinygettext/language.cpp
+++ b/src/tinygettext/language.cpp
@@ -213,7 +213,7 @@ static const LanguageSpec languages[] = {
   { "pl", "PL", 0, "Polish (Poland)"             },
   { "ps", 0,    0, "Pashto"                      },
   { "pt", 0,    0, "Portuguese"                  },
-  { "pt", "BR", 0, "Brazilian"                   },
+  { "pt", "BR", 0, "Portuguese (Brazil)"         },
   { "pt", "PT", 0, "Portuguese (Portugal)"       },
   { "qu", 0,    0, "Quechua"                     },
   { "rm", 0,    0, "Rhaeto-Romance"              },

Fix 64 bit compilation:

--- a/src/tinygettext/po_parser.cpp
+++ b/src/tinygettext/po_parser.cpp
@@ -207,7 +207,7 @@ next:
         if (pedantic)
           warning("leading whitespace before string");

-      get_string_line(out, i);
+      get_string_line(out, (unsigned int) i);
       goto next;
     }
     else if (isspace(current_line[i]))
@@ -245,7 +245,7 @@ POParser::parse_header(const std::string& header)
       if (has_prefix(line, "Content-Type:"))
       {
         // from_charset = line.substr(len);
-        unsigned int len = strlen("Content-Type: text/plain; charset=");
+        unsigned int len = (unsigned int) strlen("Content-Type: text/plain; charset=");
         if (line.compare(0, len, "Content-Type: text/plain; charset=") == 0)
         {
           from_charset = line.substr(len);

tinygettext const fix

This moves ~8kb of writable data to RO, and improves optimization. There is no need for the initial array to be writable anyhow.

--- a/src/tinygettext/language.cpp
+++ b/src/tinygettext/language.cpp
@@ -39,7 +39,7 @@ struct LanguageSpec {

 /** Language Definitions */
 //*{
-LanguageSpec languages[] = {
+static const LanguageSpec languages[] = {
   { "aa", 0,    0, "Afar"                        },
   { "af", 0,    0, "Afrikaans"                   },
   { "af", "ZA", 0, "Afrikaans (South Africa)"    },
@@ -364,7 +364,7 @@ resolve_language_alias(const std::string& name)
 Language
 Language::from_spec(const std::string& language, const std::string& country, const std::string& modifier)
 {
-  static std::map<std::string, std::vector<LanguageSpec*> > language_map;
+  static std::map<std::string, std::vector<const LanguageSpec*> > language_map;

   if (language_map.empty())
   { // Init language_map
@@ -372,10 +372,10 @@ Language::from_spec(const std::string& language, const std::string& country, con
       language_map[languages[i].language].push_back(&languages[i]);
   }

-  std::map<std::string, std::vector<LanguageSpec*> >::iterator i = language_map.find(language);
+  std::map<std::string, std::vector<const LanguageSpec*> >::iterator i = language_map.find(language);
   if (i != language_map.end())
   {
-    std::vector<LanguageSpec*>& lst = i->second;
+    std::vector<const LanguageSpec*>& lst = i->second;

     LanguageSpec tmpspec;
     tmpspec.language = language.c_str();
@@ -383,9 +383,9 @@ Language::from_spec(const std::string& language, const std::string& country, con
     tmpspec.modifier = modifier.c_str();
     Language tmplang(&tmpspec);

-    LanguageSpec* best_match = 0;
+    const LanguageSpec* best_match = 0;
     int best_match_score = 0;
-    for(std::vector<LanguageSpec*>::iterator j = lst.begin(); j != lst.end(); ++j)
+    for(std::vector<const LanguageSpec*>::iterator j = lst.begin(); j != lst.end(); ++j)
     { // Search for the language that best matches the given spec, value country more then modifier
       int score = Language::match(Language(*j), tmplang);

@@ -445,7 +445,7 @@ Language::from_env(const std::string& env)
   return from_spec(language, country, modifier);
 }

-Language::Language(LanguageSpec* language_spec_)
+Language::Language(const LanguageSpec* language_spec_)
   : language_spec(language_spec_)
 {
 }
diff --git a/src/tinygettext/language.hpp b/src/tinygettext/language.hpp
index bd4a3ea..0ab2f32 100644
--- a/src/tinygettext/language.hpp
+++ b/src/tinygettext/language.hpp
@@ -28,9 +28,9 @@ struct LanguageSpec;
 class Language
 {
 private:
-  LanguageSpec* language_spec;
+  const LanguageSpec* language_spec;

-  Language(LanguageSpec* language_spec);
+  Language(const LanguageSpec* language_spec);

 public:
   /** Create a language from language and country code:

stem() results in locale/da.po cecomming da which fails the .po sufix check.

https://github.com/tinygettext/tinygettext/blob/0c31494e2f66686cddb4c399017bb2993ff3a4bb/src/unix_file_system.cpp#L38

A workaround is to have an empty file called da.po.po

https://en.cppreference.com/w/cpp/filesystem/path/stem

Based on a pull request on our game (https://github.com/abunchofhacks/Epicinium/pull/4):

DragonFly, FreeBSD, NetBSD (but not OpenBSD) have iconv* support from Citrus Project. What causes ambiguity is system iconv_t defined as a struct pointer instead a void pointer.

$ c++ --version
FreeBSD clang version 11.0.0 ([email protected]:llvm/llvm-project.git llvmorg-11.0.0-0-g176249bd673)
Target: x86_64-unknown-freebsd13.0
Thread model: posix
InstalledDir: /usr/bin

$ CPATH=/usr/local/include gmake .obj/libs/tinygettext/iconv.o
c++ -std=c++17 -MT .obj/libs/tinygettext/iconv.o -MMD -MP -MF .dep/libs/tinygettext/iconv.Td -O3 -s -I ./ -I libs -I libs/SDL2 -o .obj/libs/tinygettext/iconv.o -c libs/tinygettext/iconv.cpp
libs/tinygettext/iconv.cpp:50:5: error: call to 'iconv_close' is ambiguous
    iconv_close(cd);
    ^~~~~~~~~~~
/usr/include/iconv.h:61:5: note: candidate function
int     iconv_close(iconv_t);
        ^
libs/tinygettext/iconv.hpp:69:12: note: candidate function
inline int iconv_close(iconv_t cd)
           ^
libs/tinygettext/iconv.cpp:57:5: error: call to 'iconv_close' is ambiguous
    iconv_close(cd);
    ^~~~~~~~~~~
/usr/include/iconv.h:61:5: note: candidate function
int     iconv_close(iconv_t);
        ^
libs/tinygettext/iconv.hpp:69:12: note: candidate function
inline int iconv_close(iconv_t cd)
           ^
libs/tinygettext/iconv.cpp:119:9: error: call to 'iconv' is ambiguous
        iconv(cd, nullptr, nullptr, nullptr, nullptr); // reset state
        ^~~~~
/usr/include/iconv.h:58:8: note: candidate function
size_t  iconv(iconv_t, char ** __restrict,
        ^
libs/tinygettext/iconv.hpp:58:15: note: candidate function
inline size_t iconv(iconv_t cd,
              ^

$ g++11 --version
g++11 (FreeBSD Ports Collection) 11.0.0 20201108 (experimental)
Copyright (C) 2020 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

$ CXX=g++11 gmake .obj/libs/tinygettext/iconv.o                                                               3& iconv γ /tmp/epicinium
g++11  -std=c++17 -MT .obj/libs/tinygettext/iconv.o -MMD -MP -MF .dep/libs/tinygettext/iconv.Td -O3 -s  -I ./  -I libs -I libs/SDL2 -o .obj/libs/tinygettext/iconv.o -c libs/tinygettext/iconv.cpp
libs/tinygettext/iconv.cpp: In destructor 'tinygettext::IConv::~IConv()':
libs/tinygettext/iconv.cpp:50:19: error: call of overloaded 'iconv_close(__tag_iconv_t*&)' is ambiguous
   50 |     iconv_close(cd);
      |                   ^
In file included from libs/tinygettext/iconv.cpp:28:
libs/tinygettext/iconv.hpp:69:12: note: candidate: 'int tinygettext::iconv_close(tinygettext::iconv_t)'
   69 | inline int iconv_close(iconv_t cd)
      |            ^~~~~~~~~~~
In file included from libs/tinygettext/iconv.hpp:28,
                 from libs/tinygettext/iconv.cpp:28:
/usr/include/iconv.h:61:9: note: candidate: 'int iconv_close(iconv_t)'
   61 | int     iconv_close(iconv_t);
      |         ^~~~~~~~~~~
libs/tinygettext/iconv.cpp: In member function 'void tinygettext::IConv::set_charsets(const string&, const string&)':
libs/tinygettext/iconv.cpp:57:19: error: call of overloaded 'iconv_close(__tag_iconv_t*&)' is ambiguous
   57 |     iconv_close(cd);
      |                   ^
In file included from libs/tinygettext/iconv.cpp:28:
libs/tinygettext/iconv.hpp:69:12: note: candidate: 'int tinygettext::iconv_close(tinygettext::iconv_t)'
   69 | inline int iconv_close(iconv_t cd)
      |            ^~~~~~~~~~~
In file included from libs/tinygettext/iconv.hpp:28,
                 from libs/tinygettext/iconv.cpp:28:
/usr/include/iconv.h:61:9: note: candidate: 'int iconv_close(iconv_t)'
   61 | int     iconv_close(iconv_t);
      |         ^~~~~~~~~~~
libs/tinygettext/iconv.cpp: In member function 'std::string tinygettext::IConv::convert(const string&)':
libs/tinygettext/iconv.cpp:119:53: error: call of overloaded 'iconv(__tag_iconv_t*&, std::nullptr_t, std::nullptr_t, std::nullptr_t, std::nullptr_t)' is ambiguous
  119 |         iconv(cd, nullptr, nullptr, nullptr, nullptr); // reset state
      |                                                     ^
In file included from libs/tinygettext/iconv.cpp:28:
libs/tinygettext/iconv.hpp:58:15: note: candidate: 'size_t tinygettext::iconv(tinygettext::iconv_t, const char**, size_t*, char**, size_t*)'
   58 | inline size_t iconv(iconv_t cd,
      |               ^~~~~
In file included from libs/tinygettext/iconv.hpp:28,
                 from libs/tinygettext/iconv.cpp:28:
/usr/include/iconv.h:58:9: note: candidate: 'size_t iconv(iconv_t, char**, size_t*, char**, size_t*)'
   58 | size_t  iconv(iconv_t, char ** __restrict,
      |         ^~~~~

I think I found a bug caused by a recent rewrite.

In commit 9bb55f3d7652e6097abcf248748509fb998cb52f, an older implementation of UnixFileSystem was replaced with one based on std::filesystem. However p.path().stem() is used instead of p.path().filename(), which means UnixFileSystem::open_directory() returns a vector of stems instead of a vector of filenames. For example it returns "en_US" if there is a file named "en_US.po". This causes the DictionaryManager to ignore all of the files in the directory, because it thinks there are no files ending on "po".

If it helps: this was when I cloned commit 0c31494e2f66686cddb4c399017bb2993ff3a4bb. ~~My current workaround is checking out commit e406e999b0c37c06cf1ec429b22669c40b16bf4f instead.~~ Replacing stem() with filename() fixed this.

cc1plus: warning: -Wabi won't warn about anything [-Wabi]
cc1plus: note: -Wabi warns about differences from the most up-to-date ABI, which is also used by default
cc1plus: note: use e.g. -Wabi=11 to warn about changes from GCC 7

Not sure what was the intention behind adding -Wabi to the list, but it should probably either be dropped or updated to actually do something :)

Currently collisions for plural strings do not tell the user what actually conflicts, so first check if something actually conflicts and if yes log all plurals for the current and the new translation.

See http://trac.wildfiregames.com/changeset/15179/ and http://trac.wildfiregames.com/ticket/2575 for the first commit.

http://trac.wildfiregames.com/changeset/15493/ for the second.

po_parser.cpp

code with truncation error: // skip UTF-8 intro that some text editors produce // see http://en.wikipedia.org/wiki/Byte-order_mark if (current_line.size() >= 3 && current_line[0] == static_cast(0xef) && current_line[1] == static_cast(0xbb) && current_line[2] == static_cast(0xbf)) { current_line = current_line.substr(3); }

can be corrected so: // skip UTF-8 intro that some text editors produce // see http://en.wikipedia.org/wiki/Byte-order_mark if (current_line.size() >= 3 && current_line[0] == '\xEF' && current_line[1] == '\xBB' && current_line[2] == '\xBF') { current_line = current_line.substr(3); }

It seems macOS doesn't define filesystem prior to macOS 10.15

builderr-debug-macos.txt
../../../source/third_party/tinygettext/src/unix_file_system.cpp:24:10: fatal error: 'filesystem' file not found
#include <filesystem>
         ^~~~~~~~~~~~
1 error generated.
make[1]: *** [obj/tinygettext_Debug/unix_file_system.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make: *** [tinygettext] Error 2

According to https://stackoverflow.com/questions/42633477/macos-clang-c17-filesystem-header-not-found we need to include <experimental/filesystem>for those OSes

so maybe a check like

// Untested
#if defined(__APPLE__) && defined(__clang_major__) && __clang_major__ < 11
#include <experimental/filesystem>
#else
#include <filesystem>
#endif

Not sure what your requirements are for this library but I understand that os might be too old, feel free to close this issue then.

Hello, I have implemented an expression parser for plural forms, which will fix #21 .

After I have implemented this, I also checked the official GNU gettext implementation, they also have an expression parser for plural forms (event for the default one n!=1).

It looks like that the last plural form description in the following file is missing an ending semicolon, compare to other plural form descriptions in the same file before:

https://github.com/tinygettext/tinygettext/blob/93ceb742f6a4160e6ca0668881db14338be9b90a/src/plural_forms.cpp#L73

Is it correct?

It looks like that in PluralForms class the input plural forms description is only simply compare with a predefined list. Is it feasible to implement a simple expression parser to handle plural forms outside the predefined list?

If yes, I'd like to implement one and submit a pull request later.