// Encodes the data from infile into the ALFileStore outfile, closing both afterward
int encodeCommandEncode(FILE *infile, uint64_t infile_len, ALStore *outfile) {
const int datatypeLen = ALDatatypeGetSize(OPTIONS.datatype);
// Allocate the input buffer
uint64_t buflen;
if (infile_len > (OPTIONS.part_size_in_elem * datatypeLen))
buflen = (OPTIONS.part_size_in_elem * datatypeLen);
else
buflen = infile_len;
void *inbuf = malloc(buflen); // Input buffer
// Prepare the output buffer (encoder)
ALIndexForm index_form = OPTIONS.index_form;
if (index_form >= ALCompressedInvertedIndex) // all other compressed index based on inverted index
index_form = ALInvertedIndex;
ALEncoderConfig econfig;
ALEncoderConfigure(&econfig, OPTIONS.significant_bits, OPTIONS.datatype, index_form);
ALPartitionData partdata; // Output buffer
size_t bytesread;
int i = 0;
double s ;
encode_time = 0;
compress_time = 0;
write_time = 0;
total_time = 0;
double ss = dclock();
while (!feof(infile) && !ferror(infile) && (bytesread = fread(inbuf, 1, buflen, infile)) != 0) {
i++;
dbprintf("Encoding partition %d with size %llu...\n", i, bytesread);
s = dclock();
ALEncode(&econfig, inbuf, bytesread / datatypeLen, &partdata);
encode_time = encode_time + (dclock() - s);
s = dclock();
if (OPTIONS.index_form >= ALCompressedInvertedIndex) // all other compressed index based on inverted index
ALConvertIndexForm(&partdata.metadata, &partdata.index, OPTIONS.index_form);
compress_time = compress_time + (dclock() - s);
s = dclock();
if (ALStoreWritePartition(outfile, &partdata) != ALErrorNone) {
fprintf(stderr, "Error appending ALACRITY partition to file, aborting\n");
abort();
return 1;
}
write_time = write_time + (dclock() - s);
ALPartitionDataDestroy(&partdata); // TODO: make this automatic when encoding over an existing partition data
dbprintf("Partition %d done!\n", i);
}
total_time = dclock() - ss;
printf("[read: %9.3lf] [encode: %9.3lf] [compress: %9.3lf] [write: %9.3lf] [total: %9.3lf]\n", total_time - (encode_time + compress_time + write_time ), encode_time, compress_time, write_time, total_time);
if (ferror(infile)) {
fprintf(stderr, "Error reading from input file, aborting\n");
return 1;
}
// Cleanup
free(inbuf);
fclose(infile);
printf("Encoding complete, %llu bytes of input data successfully encoded into %llu partitions\n", infile_len, outfile->cur_partition);
dbprintf("Closing ALACRITY output file...\n");
if (ALStoreClose(outfile) != ALErrorNone) {
fprintf(stderr, "Error closing ALACRITY output file, aborting\n");
return 1;
}
return 0;
}
int main(int argc, char **argv) {
if (argc != 4) {
fprintf(stderr, "Usage: %s <input filename> <output filename base> <invert index?>\n", argv[0]);
return 1;
}
/*
uint64_t test = 0x1234567887654321;
high_order_bytes_t hi;
low_order_bytes_t lo;
SPLIT_DATUM(&test, 8, 2, hi, lo);
printf("%016llx -> %08lx:%016llx", test, hi, lo);
return 0;
*/
const char *infilename = argv[1];
const char *outfilenamebase = argv[2];
_Bool invertIndex = atoi(argv[3]) > 0;
char outindexfilename[256];
char outdatafilename[256];
strcpy(outindexfilename, outfilenamebase);
strcat(outindexfilename, invertIndex ? "-query_index.dat" : "-index.dat");
strcpy(outdatafilename, outfilenamebase);
strcat(outdatafilename, "-compressed_data.dat");
struct stat st;
stat(infilename, &st);
uint64_t numDoubles = st.st_size / sizeof(double);
void *data = malloc(numDoubles * sizeof(double));
FILE *infile = fopen(infilename, "r");
FILE *outindexfile = fopen(outindexfilename, "w");
FILE *outdatafile = fopen(outdatafilename, "w");
if (infile == NULL) {
fprintf(stderr, "Error opening input file %s\n", infilename);
return 1;
}
if (outindexfile == NULL) {
fprintf(stderr, "Error opening output file %s\n", outindexfilename);
return 1;
}
if (outdatafile == NULL) {
fprintf(stderr, "Error opening output file %s\n", outdatafilename);
return 1;
}
int rcount = fread(data, sizeof(double), numDoubles, infile);
if (rcount != numDoubles) {
fprintf(stderr, "Expected %d doubles, read %d\n", numDoubles, rcount);
return 1;
}
ALEncoderConfig config;
ALEncoderConfigure(&config, 16, DATATYPE_FLOAT64, invertIndex ? ALInvertedIndex : ALCompressionIndex);
ALPartitionData output;
ALEncode(&config, data, numDoubles, &output);
int insigbytes = output.metadata.elementSize - output.metadata.significantBits;
int wcount = fwrite(output.index, invertIndex ? sizeof(rid_t) : output.metadata.significantBits, numDoubles, outindexfile);
if (wcount != numDoubles) {
fprintf(stderr, "Expected write %d index elements, wrote %d\n", numDoubles, wcount);
return 1;
}
wcount = fwrite(output.data, insigbytes, numDoubles, outdatafile);
if (wcount != numDoubles) {
fprintf(stderr, "Expected write %d data elements, wrote %d\n", numDoubles, wcount);
return 1;
}
if (argc == 5) {
const char *compfilenamebase = argv[4];
char compindexfilename[256];
char compdatafilename[256];
strcpy(compindexfilename, compfilenamebase);
strcat(compindexfilename, invertIndex ? "-query_index.dat" : "-index.dat");
strcpy(compdatafilename, compfilenamebase);
strcat(compdatafilename, "-compressed_data.dat");
printf("Reading comparison files %s and %s\n", compdatafilename, compindexfilename);
stat(compdatafilename, &st);
uint64_t numDoubles = st.st_size / (insigbytes);
uint64_t datasize = numDoubles * insigbytes;
uint64_t indexsize = numDoubles * (invertIndex ? sizeof(rid_t) : output.metadata.significantBits);
void *compdata = malloc(datasize);
void *compindex = malloc(indexsize);
assert(compdata != NULL);
assert(compindex != NULL);
FILE *cdf = fopen(compdatafilename, "r");
FILE *cif = fopen(compindexfilename, "r");
assert(cdf);
assert(cif);
int r1 = fread(compdata, 1, datasize, cdf);
int r2 = fread(compindex, 1, indexsize, cif);
fclose(cdf);
fclose(cif);
assert(r1 == datasize);
assert(r2 == indexsize);
printf("Testing against comparison data file %s...\n", compdatafilename);
assert(memcmp(output.data, compdata, datasize) == 0);
printf("Success!\n");
printf("Testing against comparison index file %s...\n", compindexfilename);
assert(memcmp(output.index, compindex, indexsize) == 0);
printf("Success!\n");
}
fclose(infile);
fclose(outindexfile);
fclose(outdatafile);
return 0;
}
