/* maxlen 0 = read one text line */
I32
Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
{
filter_t funcp;
SV *datasv = NULL;
if (!PL_rsfp_filters)
return -1;
if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
/* Provide a default input filter to make life easy. */
/* Note that we append to the line. This is handy. */
DEBUG_P(PerlIO_printf(Perl_debug_log,
"filter_read %d: from rsfp\n", idx));
if (maxlen) {
/* Want a block */
int len ;
const int old_len = SvCUR(buf_sv);
/* ensure buf_sv is large enough */
SvGROW(buf_sv, (STRLEN)(old_len + maxlen)) ;
if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
if (PerlIO_error(PL_rsfp))
return -1; /* error */
else
return 0 ; /* end of file */
}
SvCUR_set(buf_sv, old_len + len) ;
} else {
/* Want a line */
if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
if (PerlIO_error(PL_rsfp))
return -1; /* error */
else
return 0 ; /* end of file */
}
}
return SvCUR(buf_sv);
}
/* Skip this filter slot if filter has been deleted */
if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
DEBUG_P(PerlIO_printf(Perl_debug_log,
"filter_read %d: skipped (filter deleted)\n",
idx));
return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
}
/* Get function pointer hidden within datasv */
funcp = DPTR2FPTR(filter_t, IoANY(datasv));
DEBUG_P(PerlIO_printf(Perl_debug_log,
"filter_read %d: via function %p (%s)\n",
idx, datasv, SvPV_nolen_const(datasv)));
/* Call function. The function is expected to */
/* call "FILTER_READ(idx+1, buf_sv)" first. */
/* Return: <0:error, =0:eof, >0:not eof */
return (*funcp)(aTHX_ idx, buf_sv, maxlen);
}
static int
fmm_fhmagic(PerlFMM *state, PerlIO *fhandle, char **mime_type)
{
SV *err;
unsigned char *data;
int ret = -1;
Newz(1234, data, HOWMANY + 1, unsigned char);
if (! PerlIO_read(fhandle, data, HOWMANY)) {
err = newSVpvf(
"Failed to read from handle: %s",
strerror(errno)
);
FMM_SET_ERROR(state, err);
Safefree(data);
return -1;
}
ret = fmm_bufmagic(state, &data, mime_type);
Safefree(data);
return ret;
}
SV * parse_in_chunks(char * filepath, size_t filesize) {
char *buf;
size_t bytes_read = 0;
int max_buf = 1000;
char *err_msg;
int block = BLOCK_HEADER;
int cur_event_type = 0;
int event_type = 0;
char event_block = 0;
char *brnl, *breq;
AV * data;
AV * datawrapper;
AV * events;
char *line;
char * nl = "\n";
char * eq = "=";
int rewind_pos = 0;
size_t cur_fpos = 0;
SV * pbuf;
SV * pmax_buf;
AV * HANDLERS = get_av("Opsview::Utils::NDOLogsImporter::HANDLERS", 0);
AV * INPUT_DATA_TYPE = get_av("Opsview::Utils::NDOLogsImporter::INPUT_DATA_TYPE", 0);
int init_last_pos;
int init_block;
if ( first_read ) {
if ( ! ( fh = PerlIO_open( filepath, "rb" ) ) ) {
croak("Could not open file: %s\n", strerror(errno));
}
bytes_left = filesize;
init_last_pos = prev_pos = first_read = 0;
init_block = block = BLOCK_HEADER;
} else {
init_block = block = BLOCK_EVENTS;
init_last_pos = prev_pos;
}
read_begin:
brnl = NULL;
breq = NULL;
pbuf = get_sv("Opsview::Utils::NDOLogsImporter::PARSE_BUF", 0);
pmax_buf = get_sv("Opsview::Utils::NDOLogsImporter::MAX_BUF_SIZE", 0);
buf = SvPVX(pbuf);
max_buf = SvIVX(pmax_buf);
if ( max_buf < 1024 * 1024 && ! automated_tests ) {
max_buf = 1024*1024;
SvIV_set( pmax_buf, max_buf );
SvGROW( pbuf, max_buf + 1);
SvCUR_set( pbuf, max_buf);
}
if ( bytes_left > 0 ) {
bytes_read = PerlIO_read(fh, buf + prev_pos, max_buf-prev_pos);
cur_fpos = PerlIO_tell(fh);
if ( bytes_read < 0 ) {
err_msg = strerror(errno);
PerlIO_close( fh );
croak("Could not read file: %s\n", err_msg);
}
bytes_left -= bytes_read;
events = (AV *)sv_2mortal((SV *)newAV());
rewind_pos = last_999(buf+prev_pos, bytes_read);
prev_pos = bytes_read + prev_pos - rewind_pos;
buf[prev_pos] = '\0';
// avg ratio events:file_size = 0.21%
if ( prev_pos > 1000 ) {
av_extend( events, (int)(prev_pos * 0.0021) );
}
for ( line = strtok_r(buf, nl, &brnl); line != NULL; line = strtok_r(NULL, nl, &brnl) )
{
switch(block) {
case BLOCK_HEADER:
{
if ( strEQ(line, "STARTDATADUMP") ) {
block = BLOCK_EVENTS;
}
}
break;
case BLOCK_EVENTS:
{
if ( strEQ(line, "1000") ) { /* NDO_API_ENDDATADUMP */
block = BLOCK_FOOTER;
continue;
}
cur_event_type = atoi(line);
/* ignore events we are not handling */
if ( ! av_exists(HANDLERS, cur_event_type) ) {
block = BLOCK_IGNORE_EVENT;
continue;
}
event_block = BLOCK_EVENT_STARTED;
if ( cur_event_type != event_type ) {
datawrapper = (AV *)sv_2mortal((SV *)newAV());
data = (AV *)sv_2mortal((SV *)newAV());
av_push( events, newSViv( cur_event_type ) );
av_push( datawrapper, newRV( (SV *)data ) );
av_push( events, newRV( (SV *)datawrapper ) );
event_type = cur_event_type;
} else {
data = (AV *)sv_2mortal((SV *)newAV());
av_push( datawrapper, newRV( (SV *)data ) );
}
block = BLOCK_EVENT;
}
break;
case BLOCK_EVENT:
{
if ( strEQ(line, "999") ) { /* NDO_API_ENDDATA */
block = BLOCK_EVENTS;
event_block = BLOCK_EVENT_ENDED;
} else {
char *k;
char *v;
int key;
int key_type = 0;
int v_len = 0;
k = strtok_r(line, eq, &breq);
v = strtok_r(NULL, "\0", &breq);
key = atoi(k);
/* invalid key, skip parsing */
if ( key == 0 ) {
goto remove_invalid;
}
SV ** const k_type = av_fetch(INPUT_DATA_TYPE, key, 0 );
if ( k_type ) {
key_type = SvIVx( *k_type );
}
if ( v ) {
if ( key_type & 1 ) {
v_len = ndo_unescape_buffer( v );
} else {
v_len = strlen(v);
}
}
if ( key_type & 2 ) {
AV * datanstptr;
SV ** const datanst = av_fetch(data, key, 0 );
if ( datanst ) {
datanstptr = (AV *)SvRV( *datanst );
} else {
datanstptr = (AV *)sv_2mortal((SV *)newAV());
av_store( data, key, newRV( (SV *)datanstptr ) );
}
if ( v ) {
av_push( datanstptr, newSVpvn(v, v_len) );
} else {
av_push( datanstptr, newSVpvn("", 0) );
}
} else {
if ( v ) {
av_store( data, key, newSVpvn(v, v_len) );
} else {
av_store( data, key, newSVpvn("", 0) );
}
}
}
}
break;
case BLOCK_FOOTER:
{
if ( strEQ(line, "GOODBYE") ) {
block = BLOCK_HEADER;
}
}
break;
case BLOCK_IGNORE_EVENT:
{
if ( strEQ(line, "999") ) { /* NDO_API_ENDDATA */
block = BLOCK_EVENTS; // go back to EVENTS
continue;
}
}
break;
}
};
/* there were some events */
if ( event_block != BLOCK_HEADER ) {
if ( event_block != BLOCK_EVENT_ENDED ) {
remove_invalid:
av_pop( datawrapper );
}
/* remove whole block if the last block has no events */
if ( av_len( datawrapper ) == -1 ) {
av_pop( events );
av_pop( events );
}
}
if ( av_len(events) > 0 ) {
if ( rewind_pos > 0 && cur_fpos < filesize ) {
memmove(buf, buf+prev_pos+1, rewind_pos-1);
}
prev_pos = rewind_pos - 1;
return newRV_inc((SV *) events);
} else {
if ( cur_fpos < filesize && event_block != BLOCK_HEADER && event_block != BLOCK_EVENT_ENDED ) {
int new_max_buf = max_buf * 2;
SvIV_set( pmax_buf, new_max_buf );
SvGROW( pbuf, new_max_buf + 1);
SvCUR_set( pbuf, new_max_buf);
//start again as previous buffer would be tokenized already
prev_pos = 0;
block = init_block;
event_type = 0;
PerlIO_close( fh );
if ( ! ( fh = PerlIO_open( filepath, "rb" ) ) ) {
croak("Could not re-open file: %s\n", strerror(errno));
}
PerlIO_seek(fh, cur_fpos-bytes_read-init_last_pos, SEEK_SET);
bytes_left += bytes_read + init_last_pos;
goto read_begin;
}
}
}
parser_reset_iterator();
return &PL_sv_undef;
}
int
_check_buf(PerlIO *infile, Buffer *buf, int min_wanted, int max_wanted)
{
int ret = 1;
// Do we have enough data?
if ( buffer_len(buf) < min_wanted ) {
// Read more data
uint32_t read;
uint32_t actual_wanted;
unsigned char *tmp;
#ifdef _MSC_VER
uint32_t pos_check = PerlIO_tell(infile);
#endif
if (min_wanted > max_wanted) {
max_wanted = min_wanted;
}
// Adjust actual amount to read by the amount we already have in the buffer
actual_wanted = max_wanted - buffer_len(buf);
New(0, tmp, actual_wanted, unsigned char);
DEBUG_TRACE("Buffering from file @ %d (min_wanted %d, max_wanted %d, adjusted to %d)\n",
(int)PerlIO_tell(infile), min_wanted, max_wanted, actual_wanted
);
if ( (read = PerlIO_read(infile, tmp, actual_wanted)) <= 0 ) {
if ( PerlIO_error(infile) ) {
#ifdef _MSC_VER
// Show windows specific error message as Win32 PerlIO_read does not set errno
DWORD last_error = GetLastError();
LPWSTR *errmsg = NULL;
FormatMessage(FORMAT_MESSAGE_ALLOCATE_BUFFER | FORMAT_MESSAGE_FROM_SYSTEM, 0, last_error, 0, (LPWSTR)&errmsg, 0, NULL);
warn("Error reading: %d %s (read %d wanted %d)\n", last_error, errmsg, read, actual_wanted);
LocalFree(errmsg);
#else
warn("Error reading: %s (wanted %d)\n", strerror(errno), actual_wanted);
#endif
}
else {
warn("Error: Unable to read at least %d bytes from file.\n", min_wanted);
}
ret = 0;
goto out;
}
buffer_append(buf, tmp, read);
// Make sure we got enough
if ( buffer_len(buf) < min_wanted ) {
warn("Error: Unable to read at least %d bytes from file (only read %d).\n", min_wanted, read);
ret = 0;
goto out;
}
#ifdef _MSC_VER
// Bug 16095, weird off-by-one bug seen only on Win32 and only when reading a filehandle
if (PerlIO_tell(infile) != pos_check + read) {
//PerlIO_printf(PerlIO_stderr(), "Win32 bug, pos should be %d, but was %d\n", pos_check + read, PerlIO_tell(infile));
PerlIO_seek(infile, pos_check + read, SEEK_SET);
}
#endif
DEBUG_TRACE("Buffered %d bytes, new pos %d\n", read, (int)PerlIO_tell(infile));
out:
Safefree(tmp);
}