/* PSPP - a program for statistical analysis.
- Copyright (C) 2006, 2009, 2010, 2011, 2012, 2013 Free Software Foundation, Inc.
+ Copyright (C) 2006, 2009, 2010, 2011, 2012, 2013, 2014 Free Software Foundation, Inc.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
#include <errno.h>
#include <iconv.h>
#include <langinfo.h>
-#include <libintl.h>
#include <locale.h>
#include <stdio.h>
#include <stdlib.h>
Returns the output length if successful, -1 if the output buffer is too
small. */
static ssize_t
-try_recode (iconv_t conv,
- const char *ip, size_t inbytes,
- char *op_, size_t outbytes)
+try_recode (iconv_t conv, char fallbackchar,
+ const char *in, size_t inbytes,
+ char *out_, size_t outbytes)
{
- /* FIXME: Need to ensure that this char is valid in the target encoding */
- const char fallbackchar = '?';
- char *op = op_;
+ char *out = out_;
+ int i;
/* Put the converter into the initial shift state, in case there was any
state information left over from its last usage. */
iconv (conv, NULL, 0, NULL, 0);
- while (iconv (conv, (ICONV_CONST char **) &ip, &inbytes,
- &op, &outbytes) == -1)
- switch (errno)
- {
- case EINVAL:
- if (outbytes < 2)
- return -1;
- *op++ = fallbackchar;
- *op = '\0';
- return op - op_;
-
- case EILSEQ:
- if (outbytes == 0)
- return -1;
- *op++ = fallbackchar;
- outbytes--;
- ip++;
- inbytes--;
- break;
-
- case E2BIG:
- return -1;
-
- default:
- /* should never happen */
- fprintf (stderr, "Character conversion error: %s\n", strerror (errno));
- NOT_REACHED ();
- break;
- }
+ /* Do two rounds of iconv() calls:
+
+ - The first round does the bulk of the conversion using the
+ caller-supplied input data..
+
+ - The second round flushes any leftover output. This has a real effect
+ with input encodings that use combining diacritics, e.g. without the
+ second round the last character tends to gets dropped when converting
+ from windows-1258 to other encodings.
+ */
+ for (i = 0; i < 2; i++)
+ {
+ ICONV_CONST char **inp = i ? NULL : (ICONV_CONST char **) ∈
+ size_t *inbytesp = i ? NULL : &inbytes;
+
+ while (iconv (conv, inp, inbytesp, &out, &outbytes) == -1)
+ switch (errno)
+ {
+ case EINVAL:
+ if (outbytes < 2)
+ return -E2BIG;
+ if (!fallbackchar)
+ return -EINVAL;
+ *out++ = fallbackchar;
+ *out = '\0';
+ return out - out_;
+
+ case EILSEQ:
+ if (outbytes == 0)
+ return -E2BIG;
+ if (!fallbackchar)
+ return -EILSEQ;
+ *out++ = fallbackchar;
+ outbytes--;
+ if (inp)
+ {
+ in++;
+ inbytes--;
+ }
+ break;
+
+ case E2BIG:
+ return -E2BIG;
+
+ default:
+ /* should never happen */
+ fprintf (stderr, "Character conversion error: %s\n",
+ strerror (errno));
+ NOT_REACHED ();
+ break;
+ }
+ }
if (outbytes == 0)
- return -1;
+ return -E2BIG;
- *op = '\0';
- return op - op_;
+ *out = '\0';
+ return out - out_;
}
/* Converts the string TEXT, which should be encoded in FROM-encoding, to a
return recode_string ("UTF-8", filename_encoding (), filename, -1);
}
+static int
+recode_substring_pool__ (const char *to, const char *from,
+ struct substring text, char fallbackchar,
+ struct pool *pool, struct substring *out)
+{
+ size_t bufsize;
+ iconv_t conv ;
+
+ if (to == NULL)
+ to = default_encoding;
+
+ if (from == NULL)
+ from = default_encoding;
+
+ conv = create_iconv (to, from);
+
+ if ( (iconv_t) -1 == conv )
+ {
+ if (fallbackchar)
+ {
+ out->string = pool_malloc (pool, text.length + 1);
+ out->length = text.length;
+ memcpy (out->string, text.string, text.length);
+ out->string[out->length] = '\0';
+ return 0;
+ }
+ else
+ return EPROTO;
+ }
+
+ for (bufsize = text.length + 1; bufsize > text.length; bufsize *= 2)
+ {
+ char *output = pool_malloc (pool, bufsize);
+ ssize_t retval;
+
+ retval = try_recode (conv, fallbackchar, text.string, text.length,
+ output, bufsize);
+ if (retval >= 0)
+ {
+ *out = ss_buffer (output, retval);
+ return 0;
+ }
+ pool_free (pool, output);
+
+ if (retval != -E2BIG)
+ return -retval;
+ }
+
+ NOT_REACHED ();
+}
+
/* Converts the string TEXT, which should be encoded in FROM-encoding, to a
dynamically allocated string in TO-encoding. Any characters which cannot be
converted will be represented by '?'.
recode_substring_pool (const char *to, const char *from,
struct substring text, struct pool *pool)
{
- size_t outbufferlength;
- iconv_t conv ;
-
- if (to == NULL)
- to = default_encoding;
-
- if (from == NULL)
- from = default_encoding;
-
- conv = create_iconv (to, from);
+ struct substring out;
- if ( (iconv_t) -1 == conv )
- {
- struct substring out;
+ recode_substring_pool__ (to, from, text, '?', pool, &out);
+ return out;
+}
- out.string = pool_malloc (pool, text.length + 1);
- out.length = text.length;
- memcpy (out.string, text.string, text.length);
- out.string[out.length] = '\0';
- return out;
- }
+/* Converts the string TEXT, which should be encoded in FROM-encoding, to a
+ dynamically allocated string in TO-encoding. On success, returns 0, and the
+ converted null-terminated string, allocated from POOL with pool_malloc(), is
+ stored in *OUT. On failure, returns a positive errno value.
- for ( outbufferlength = 1 ; outbufferlength != 0; outbufferlength <<= 1 )
- if ( outbufferlength > text.length)
- {
- char *output = pool_malloc (pool, outbufferlength);
- ssize_t output_len = try_recode (conv, text.string, text.length,
- output, outbufferlength);
- if (output_len >= 0)
- return ss_buffer (output, output_len);
- pool_free (pool, output);
- }
+ The function fails with an error if any part of the input string is not
+ valid in the declared input encoding. */
+int
+recode_pedantically (const char *to, const char *from,
+ struct substring text, struct pool *pool,
+ struct substring *out)
+{
+ int error;
- NOT_REACHED ();
+ error = recode_substring_pool__ (to, from, text, 0, pool, out);
+ if (error)
+ *out = ss_empty ();
+ return error;
}
-
+\f
void
i18n_init (void)
{
projects / pspp / blobdiff