[Refdb-cvs] CVS: refdb/src refdbdref.c,1.51,1.52

[Markus H. <mho...@us...>]

Update of /cvsroot/refdb/refdb/src
In directory sc8-pr-cvs1:/tmp/cvs-serv12453

Modified Files:
	refdbdref.c 
Log Message:
do character conversion for RIS data if required

Index: refdbdref.c
===================================================================
RCS file: /cvsroot/refdb/refdb/src/refdbdref.c,v
retrieving revision 1.51
retrieving revision 1.52
diff -u -U2 -r1.51 -r1.52
--- refdbdref.c	17 Jan 2004 23:07:41 -0000	1.51
+++ refdbdref.c	18 Jan 2004 01:01:02 -0000	1.52
@@ -336,12 +336,14 @@
 
   /* if we need to convert, create a conversion descriptor for iconv() */
-  if (db_encoding && strcmp(db_encoding, "UTF-8")) {
-    ardata.conv_descriptor = iconv_open(!strcmp(db_encoding, "US-ASCII") ? "ASCII" : db_encoding, "UTF-8");
+  if (db_encoding && strcmp(db_encoding, !strcmp(reftype, "risx") ? "UTF-8" : input_encoding)) {
+    ardata.conv_descriptor = iconv_open(!strcmp(db_encoding, "US-ASCII") ? "ASCII" : db_encoding, !strcmp(reftype, "risx") ? "UTF-8" : input_encoding);
     if (ardata.conv_descriptor == (iconv_t)(-1)) {
       ardata.conv_descriptor = NULL;
-      LOG_PRINT(LOG_WARNING, "cannot set output encoding");
+      LOG_PRINT(LOG_WARNING, "cannot set conversion descriptor");
     }
     else {
-      LOG_PRINT(LOG_DEBUG, "encoding is now:");
+      LOG_PRINT(LOG_DEBUG, "input encoding is:");
+      LOG_PRINT(LOG_DEBUG, input_encoding);
+      LOG_PRINT(LOG_DEBUG, "output encoding is:");
       LOG_PRINT(LOG_DEBUG, db_encoding);
     }
@@ -349,5 +351,5 @@
   else {
     ardata.conv_descriptor = NULL;
-    LOG_PRINT(LOG_DEBUG, "encoding is UTF-8, no conversion required");
+    LOG_PRINT(LOG_DEBUG, "no character encoding conversion required");
   }
 
@@ -590,4 +592,10 @@
     /* read the data proper unless size was zero (= end of transmission) */
     if (n_requested_bufsize) {
+      size_t inlength;
+      size_t outlength;
+      char* my_ris_set = NULL; /* this ptr will be modified by iconv() */
+      char* my_ris_set_start = NULL; /* records initial state of my_elvalue */
+      const char* my_instring = NULL; /* this ptr will be modified by iconv() */
+
       /* acknowledge that we have a buffer ready, let them data come... */
       numbyte = iwrite(ptr_clrequest->fd, positive.text, positive.length+1);
@@ -609,4 +617,60 @@
       /* ToDo: run iconv here */
 
+      if (conv_descriptor && *ris_set) {
+	inlength = strlen(ris_set) + 1;
+	/* with the encodings supported by our database engines, the converted
+	   string can't be longer than the input string */
+	outlength = inlength;
+
+	if ((my_ris_set = (char*)malloc(outlength)) == NULL) {
+	  ptr_addresult->failure++;
+	  sprintf(return_msg, "out of memory\n");
+	  LOG_PRINT(LOG_WARNING, "failed processing dataset");
+	  numbyte = iwrite(ptr_clrequest->fd, return_msg, strlen(return_msg)+1);
+	  if (numbyte == -1) {
+	    LOG_PRINT(LOG_INFO, "timeout while writing");
+	    free(ris_set);
+	    return 0;
+	  }
+	}
+
+	/* keep start of the converted string */
+	my_ris_set_start = my_ris_set;
+
+	/* variable will be modified by iconv, so don't use original */
+	my_instring = (const char*)ris_set;
+
+	/* now actually do the conversion */
+	if (iconv(conv_descriptor, &my_instring, &inlength, &my_ris_set, &outlength) == (size_t)(-1)) {
+	  if (errno == EILSEQ) {
+	    sprintf(return_msg, "iconv: invalid input character sequence\n");
+	    LOG_PRINT(LOG_WARNING, "iconv: invalid input character sequence");
+	  }
+	  else if (errno == E2BIG) {
+	    sprintf(return_msg, "iconv: output buffer too small\n");
+	    LOG_PRINT(LOG_WARNING, "iconv: output buffer too small");
+	  }
+	  else if (errno == EINVAL) {
+	    sprintf(return_msg, "iconv: incomplete input character\n");
+	    LOG_PRINT(LOG_WARNING, "iconv: incomplete input character");
+	  }
+	  
+	  numbyte = iwrite(ptr_clrequest->fd, return_msg, strlen(return_msg)+1);
+	  if (numbyte == -1) {
+	    LOG_PRINT(LOG_INFO, "timeout while writing");
+	    free(ris_set);
+	    return 0;
+	  }
+	  return 0;
+	}
+	/* else: conversion went ok. We free the original string and replace
+	   it with the converted copy */
+	if (ris_set) {
+	  free(ris_set);
+	}
+	ris_set = my_ris_set_start;
+      }
+      /* else: no conversion required */
+
       /* split reference into lines and squeeze contents into database */
       result = process_ris_set(ris_set, conn, conn_refdb, replace_ref, (char*)the_user, ptr_clrequest, n_keep_id, ptr_sentinel);