bool LDAPExpr::CompareString( const std::string& s1, int op, const std::string& s2 )
{
switch(op)
{
case LE:
return s1.compare(s2) <= 0;
case GE:
return s1.compare(s2) >= 0;
case EQ:
return PatSubstr(s1,s2);
case APPROX:
return FixupString(s2) == FixupString(s1);
default:
return false;
}
}
SMCError TextParsers::ParseStream_SMC(void *stream,
STREAMREADER srdr,
ITextListener_SMC *smc,
SMCStates *pStates)
{
char *reparse_point = NULL;
char in_buf[4096];
char *parse_point = in_buf;
char *line_begin = in_buf;
unsigned int read;
unsigned int curlevel = 0;
bool in_quote = false;
bool ignoring = false;
bool eol_comment = false;
bool ml_comment = false;
unsigned int i;
SMCError err = SMCError_Okay;
SMCResult res;
SMCStates states;
char c;
StringInfo strings[3];
StringInfo emptystring;
states.line = 1;
states.col = 0;
smc->ReadSMC_ParseStart();
/**
* The stream reader reads in as much as it can fill the buffer with.
* It then processes the buffer. If the buffer cannot be fully processed, for example,
* a line is left hanging with no newline, then the contents of the buffer is shifted
* down, and the buffer is filled from the stream reader again.
*
* What makes this particularly annoying is that we cache pointers everywhere, so when
* the shifting process takes place, all those pointers must be shifted as well.
*/
while (srdr(stream, parse_point, sizeof(in_buf)-(parse_point - in_buf) - 1, &read))
{
if (!read)
{
break;
}
/* Check for BOM markings, which is only relevant on the first line.
* Not worth it, but it could be moved out of the loop.
*/
if (states.line == 1 &&
in_buf[0] == (char)0xEF &&
in_buf[1] == (char)0xBB &&
in_buf[2] == (char)0xBF)
{
/* Move EVERYTHING down :\ */
memmove(in_buf, &in_buf[3], read - 3);
read -= 3;
}
if (reparse_point)
{
read += (parse_point - reparse_point);
parse_point = reparse_point;
reparse_point = NULL;
}
for (i = 0; i<read; i++)
{
c = parse_point[i];
if (c == '\n')
{
/* If we got a newline, there's a lot of things that could have happened in the interim.
* First, let's make sure the staged strings are rotated.
*/
if (strings[0].ptr)
{
strings[0].end = &parse_point[i];
if (rotate(strings) != NULL)
{
err = SMCError_InvalidTokens;
goto failed;
}
}
/* Next, let's clear some line-based values that may no longer have meaning */
eol_comment = false;
in_quote = false;
if (ignoring && !ml_comment)
{
ignoring = false;
}
/* Pass the raw line onto the listener. We terminate the line so the receiver
* doesn't get tons of useless info. We restore the newline after.
*/
parse_point[i] = '\0';
if ((res = smc->ReadSMC_RawLine(&states, line_begin)) != SMCResult_Continue)
{
err = (res == SMCResult_HaltFail) ? SMCError_Custom : SMCError_Okay;
goto failed;
}
parse_point[i] = '\n';
/* Now we check the sanity of our staged strings! */
if (strings[2].ptr)
{
if (!curlevel)
{
err = SMCError_InvalidProperty1;
goto failed;
}
/* Assume the next string is a property and pass the info on. */
if ((res = smc->ReadSMC_KeyValue(
&states,
FixupString(strings[2]),
FixupString(strings[1]))) != SMCResult_Continue)
{
err = (res == SMCResult_HaltFail) ? SMCError_Custom : SMCError_Okay;
goto failed;
}
scrap(strings);
}
/* Change the states for the next line */
states.col = 0;
states.line++;
line_begin = &parse_point[i + 1]; //Note: safe because this gets relocated later
}
else if (ignoring)
{
if (in_quote)
{
/* If i was 0, we could have reparsed, so make sure there's no buffer underrun */
if ((&parse_point[i] != in_buf) && c == '"' && parse_point[i - 1] != '\\')
{
/* If we reached a quote in an ignore phase,
* we're staging a string and we must rotate it out.
*/
in_quote = false;
ignoring = false;
/* Set our info */
strings[0].end = &parse_point[i];
strings[0].quoted = true;
if (rotate(strings) != NULL)
{
/* If we rotated too many strings, there was too much crap on one line */
err = SMCError_InvalidTokens;
goto failed;
}
}
else if (c == '\\')
{
strings[0].special = true;
if (i == (read - 1))
{
reparse_point = &parse_point[i];
break;
}
}
}
else if (ml_comment)
{
if (c == '*')
{
/* Check if we need to get more input first */
if (i == read - 1)
{
reparse_point = &parse_point[i];
break;
}
if (parse_point[i + 1] == '/')
{
ml_comment = false;
ignoring = false;
/* We should not be staging anything right now. */
assert(strings[0].ptr == NULL);
/* Advance the input stream so we don't choke on this token */
i++;
states.col++;
}
}
}
}
else
{
/* Check if we're whitespace or not */
if (!g_ws_chartable[(unsigned char)c])
{
bool restage = false;
/* Check various special tokens:
* ;
* //
* / *
* {
* }
*/
if (c == ';' || c == '/')
{
/* If it's a line-based comment (that is, ; or //)
* we will need to scrap everything until the end of the line.
*/
if (c == '/')
{
if (i == read - 1)
{
/* If we reached the end of the look-ahead, we need to re-check our input.
* Breaking out will force this to be the new reparse point!
*/
reparse_point = &parse_point[i];
break;
}
if (parse_point[i + 1] == '/')
{
/* standard comment */
ignoring = true;
eol_comment = true;
restage = true;
}
else if (parse_point[i + 1] == '*')
{
/* inline comment - start ignoring */
ignoring = true;
ml_comment = true;
/* yes, we restage, meaning that:
* STR/ *stuff* /ING (space because ml comments don't nest in C++)
* will not generate 'STRING', but rather 'STR' and 'ING'.
* This should be a rare occurrence and is done here for convenience.
*/
restage = true;
}
}
else
{
ignoring = true;
eol_comment = true;
restage = true;
}
}
else if (c == '{')
{
/* If we are staging a string, we must rotate here */
if (strings[0].ptr)
{
/* We have unacceptable tokens on this line */
if (rotate(strings) != NULL)
{
err = SMCError_InvalidSection1;
goto failed;
}
}
/* Sections must always be alone */
if (strings[2].ptr != NULL)
{
err = SMCError_InvalidSection1;
goto failed;
}
else if (strings[1].ptr == NULL)
{
err = SMCError_InvalidSection2;
goto failed;
}
if ((res = smc->ReadSMC_NewSection(&states, FixupString(strings[1])))
!= SMCResult_Continue)
{
err = (res == SMCResult_HaltFail) ? SMCError_Custom : SMCError_Okay;
goto failed;
}
strings[1] = emptystring;
curlevel++;
}
else if (c == '}')
{
/* Unlike our matching friend, this can be on the same line as something prior */
if (rotate(strings) != NULL)
{
err = SMCError_InvalidSection3;
goto failed;
}
if (strings[2].ptr)
{
if (!curlevel)
{
err = SMCError_InvalidProperty1;
goto failed;
}
if ((res = smc->ReadSMC_KeyValue(
&states,
FixupString(strings[2]),
FixupString(strings[1])))
!= SMCResult_Continue)
{
err = (res == SMCResult_HaltFail) ? SMCError_Custom : SMCError_Okay;
goto failed;
}
}
else if (strings[1].ptr)
{
err = SMCError_InvalidSection3;
goto failed;
}
else if (!curlevel)
{
err = SMCError_InvalidSection4;
goto failed;
}
/* Now it's safe to leave the section */
scrap(strings);
if ((res = smc->ReadSMC_LeavingSection(&states)) != SMCResult_Continue)
{
err = (res == SMCResult_HaltFail) ? SMCError_Custom : SMCError_Okay;
goto failed;
}
curlevel--;
}
else if (c == '"')
{
/* If we get a quote mark, we always restage, but we need to do it beforehand */
if (strings[0].ptr)
{
strings[0].end = &parse_point[i];
if (rotate(strings) != NULL)
{
err = SMCError_InvalidTokens;
goto failed;
}
}
strings[0].ptr = &parse_point[i];
in_quote = true;
ignoring = true;
}
else if (!strings[0].ptr)
{
/* If we have no string, we must start one */
strings[0].ptr = &parse_point[i];
}
if (restage && strings[0].ptr)
{
strings[0].end = &parse_point[i];
if (rotate(strings) != NULL)
{
err = SMCError_InvalidTokens;
goto failed;
}
}
}
else
{
/* If we're eating a string and get whitespace, we need to restage.
* (Note that if we are quoted, this is being ignored)
*/
if (strings[0].ptr)
{
/*
* The specification says the second string in a pair does not need to be quoted.
* Thus, we check if there's already a string on the stack.
* If there's a newline, we always rotate so the newline has an empty starter.
*/
if (!strings[1].ptr)
{
/* There's no string, so we must move this one down and eat up another */
strings[0].end = &parse_point[i];
rotate(strings);
}
else if (!strings[1].quoted)
{
err = SMCError_InvalidTokens;
goto failed;
}
}
}
}
/* Advance which token we're on */
states.col++;
}
if (line_begin != in_buf)
{
/* The line buffer has advanced, so it's safe to copy N bytes back to the beginning.
* What's N? N is the lowest point we're currently relying on.
*/
char *stage = lowstring(strings);
if (!stage || stage > line_begin)
{
stage = line_begin;
}
unsigned int bytes = read - (stage - parse_point);
/* It is now safe to delete everything before the staged point */
memmove(in_buf, stage, bytes);
/* Calculate the number of bytes in the new buffer */
bytes = stage - in_buf;
/* Relocate all the cached pointers to our new base */
line_begin -= bytes;
reloc(strings[0], bytes);
reloc(strings[1], bytes);
reloc(strings[2], bytes);
if (reparse_point)
{
reparse_point -= bytes;
}
if (parse_point)
{
parse_point = &parse_point[read];
parse_point -= bytes;
}
}
else if (read == sizeof(in_buf)-1)
{
err = SMCError_TokenOverflow;
goto failed;
}
}
/* If we're done parsing and there are tokens left over... */
if (curlevel)
{
err = SMCError_InvalidSection5;
goto failed;
}
else if (strings[0].ptr || strings[1].ptr)
{
err = SMCError_InvalidTokens;
goto failed;
}
smc->ReadSMC_ParseEnd(false, false);
if (pStates != NULL)
{
*pStates = states;
}
return SMCError_Okay;
failed:
if (pStates != NULL)
{
*pStates = states;
}
smc->ReadSMC_ParseEnd(true, (err == SMCError_Custom));
return err;
}
