void coin::WordCounter::write_frequencies(const string& file, int minimum) const{
ofstream out(file, ios::out|ios::binary|ios::trunc);
double total = saving_->total();
out.write((char *)&total, sizeof(double));
Bucket<string> *bucket = saving_->least_bucket();
int64_t counter = 0;
do{
bucket = bucket->prev();
if(bucket->count() < minimum){
bucket = bucket->next();
break;
}
Element<string> *child = bucket->child();
map<string, Element<string>*> children;
do{
children[child->item()] = child;
child = child->next();
++counter;
}while(child != bucket->child());
int childs = children.size();
out.write((char *)&childs, sizeof(int));
double count = bucket->count();
out.write((char *)&count, sizeof(double));
for(pair<string, Element<string>*> child:children){
short size = child.first.size();
out.write((char *)&size, sizeof(short));
out.write(child.first.c_str(), sizeof(char) * size);
}
if(counter > saving_->max_count()){
break;
}
}while(bucket != saving_->least_bucket());
for(pair<string, double> dict_count:dict_counts_){
int childs = 1;
out.write((char *)&childs, sizeof(int));
out.write((char *)&dict_count.second, sizeof(double));
short size = dict_count.first.size();
out.write((char *)&size, sizeof(short));
out.write(dict_count.first.c_str(), sizeof(char) * size);
}
out.close();
cerr << counter << " words with at least " << bucket->count() << " frequencies are written" << endl;
}
void func()
{
// test when file does not exist
{
bool ok = false;
try
{
StatFile sf("doesnotexist.sf", Stats(), mm);
}
catch (std::runtime_error const &re)
{
ok = true;
}
assert_true(ok);
}
setUp();
// test mapTimeToBucketIndex() and mapBucketIndexToFileIndex
{
StatFile sf("created.sf", Stats(), mm);
assert_equal(2, sf.lastBucket());
assert_equal(0, sf.mapTimeToBucketIndex(1000000000));
assert_equal(0, sf.mapTimeToBucketIndex(1000000001));
assert_equal(0, sf.mapBucketIndexToFileIndex(0));
assert_equal(1, sf.mapTimeToBucketIndex(1000000010));
assert_equal(1, sf.mapBucketIndexToFileIndex(1));
assert_equal(2, sf.mapTimeToBucketIndex(1000000020));
assert_equal(2, sf.mapBucketIndexToFileIndex(2));
int64_t calculated_first_bucket_time = 1000000020 - (10240-1)*10;
assert_equal(-10237, sf.mapTimeToBucketIndex(calculated_first_bucket_time));
assert_equal(3, sf.mapBucketIndexToFileIndex(-10237));
// go backwards across the wrap boundry
assert_equal(-1, sf.mapTimeToBucketIndex(1000000000-10));
assert_equal(10239, sf.mapBucketIndexToFileIndex(-1));
// go forwards twice across the wrap boundry
assert_equal(10241, sf.mapTimeToBucketIndex(1000000020+(10239*10)));
assert_equal(-1, sf.mapBucketIndexToFileIndex(10241));
// go forwards twice across the wrap boundry
assert_equal(20481, sf.mapTimeToBucketIndex(1000000020+(10240*10)+(10239*10)));
assert_equal(-1, sf.mapBucketIndexToFileIndex(20481));
// test round_up
assert_equal(1, sf.mapTimeToBucketIndex(1000000001, true));
assert_equal(1, sf.mapBucketIndexToFileIndex(1));
}
mm->dispose();
mm = NewMmap();
// test numBucketsBetween()
{
StatFile sf("created.sf", Stats(), mm);
assert_equal(2, sf.lastBucket());
assert_equal(1, sf.numBucketsBetween(1000000000, 1000000000));
assert_equal(1, sf.numBucketsBetween(1000000000, 1000000001));
assert_equal(2, sf.numBucketsBetween(1000000001, 1000000011, true));
assert_equal(1, sf.numBucketsBetween(1000000000, 1000000010));
}
mm->dispose();
mm = NewMmap();
{
StatFile sf("created.sf", Stats(), mm);
assert_equal(2, sf.lastBucket());
Bucket const &b = sf.bucket(0);
assert_equal(b.count(), 3);
assert_equal(b.min(), 4.0);
assert_equal(b.max(), 5.0);
assert_equal(b.sum(), 14.0);
assert_equal(b.sumSq(), 66);
assert_equal(b.time(), 1000000000);
Bucket const &c = sf.bucket(1);
assert_equal(c.count(), 2);
assert_equal(c.sumSq(), 16+25);
assert_equal(c.time(), 1000000010);
assert_equal(sf.settings().intervalTime, 10);
ssize_t bpp = 8192 / sizeof(Bucket);
ssize_t n = 10000;
int ns = 0;
while (n > 0)
{
ns = ns + bpp;
n = n - bpp;
}
assert_equal(sf.settings().numSamples, ns);
assert_equal(sf.settings().unit[0], 0);
}
mm->dispose();
mm = NewMmap();
// test that pre-fetching is turned off
{
WrapMmap wm(mm);
StatFile sf("created.sf", Stats(), &wm);
assert_equal(2, sf.lastBucket());
Bucket const &b = sf.bucket(2);
assert_equal(b.count(), 3);
assert_equal(wm.open_, 1);
assert_equal(wm.close_, 0);
assert_equal(wm.map_, 2); // file header, first page
assert_equal(wm.unmap_, 0);
assert_equal(wm.lastOffsetMapped_, 8192); // page size
size_t n = 8192 / sizeof(Bucket);
for (size_t i = 3; i < n-1; ++i)
{
Bucket c(10, 50, 5, 5, 2, 1000000000 + 10 * i);
sf.updateBucket(c);
}
assert_equal(wm.map_, 2);
assert_equal(wm.lastOffsetMapped_, 8192);
{
Bucket c(10, 50, 5, 5, 2, 1000000000 + 10 * n);
sf.updateBucket(c);
}
// fetched new now that we hit the next bucket
assert_equal(wm.map_, 3);
assert_equal(wm.lastOffsetMapped_, 8192 * 2);
int64_t no, nc, nm, nu;
mm->counters(&nm, &nu, &no, &nc);
assert_equal(nm, wm.map_);
assert_equal(nu, wm.unmap_);
assert_equal(no, wm.open_);
assert_equal(nc, wm.close_);
}
// test that we deal with intermediate empty buckets
{
setUp();
StatFile sf("created.sf", Stats(), mm);
Bucket b(123, 456, 3323, 4, 3, 1000000000 + 100);
sf.updateBucket(b);
for (int i = 0; i < 9; ++i)
{
Bucket const &c = sf.bucket(i);
if (c.time() != 0)
{
assert_equal(c.time(), 1000000000 + i * 10);
}
}
Bucket const &c = sf.bucket(6);
assert_equal(c.min(), 0);
Bucket const &d = sf.bucket(10);
assert_equal(d.min(), 3323);
}
// test seasonal buckets
{
unlink("season.sf");
StatFile::Settings settings;
settings.zeroTime = 1000000000;
settings.intervalTime = 10;
settings.numSamples = 8640;
settings.flags = FILE_FLAG_IS_TRAILING;
settings.season = 86400;
settings.lambda = 0.5;
StatFile sf("season.sf", Stats(), settings, mm, true);
{
Bucket b1(2, 4, 2, 2, 1, 1000100000);
sf.updateBucket(b1);
sf.updateBucket(Bucket(b1, 1000100100));
sf.updateBucket(Bucket(b1, 1000100110));
sf.updateBucket(Bucket(b1, 1000100120));
sf.updateBucket(Bucket(b1, 1000100130));
sf.updateBucket(Bucket(b1, 1000100140));
sf.updateBucket(Bucket(b1, 1000110100));
sf.updateBucket(Bucket(b1, 1000110110));
sf.updateBucket(Bucket(b1, 1000110120));
sf.updateBucket(Bucket(b1, 1000110130));
sf.updateBucket(Bucket(b1, 1000110140));
}
{
// 86400 after the previous write -- one season
Bucket b2(0, 0, 0, 0, 1, 1000196500);
sf.updateBucket(b2);
sf.updateBucket(Bucket(b2,1000196530));
sf.updateBucket(Bucket(b2,1000196540));
sf.updateBucket(Bucket(b2,1000196550));
}
int64_t ix = sf.mapTimeToBucketIndex(1000196500);
Bucket bOut;
int64_t r = sf.readBuckets(&bOut, 1, ix);
assert_equal(r, 1);
assert_equal(bOut.count(), 1);
assert_equal(bOut.min(), 0.5 * 2);
assert_equal(bOut.max(), 0.5 * 2);
assert_equal(bOut.sum(), 0.5 * 2);
assert_equal(bOut.sumSq(), 0.5 * 4);
sf.flush();
}
//test time_to_bucket_index returns -1 for time older than bucket range
{
setUp();
StatFile sf("created.sf", Stats(), mm);
StatFile::Settings const &settings = sf.settings();
Bucket const &last_bucket = sf.bucket(sf.lastBucket());
time_t last_time = last_bucket.time();
time_t old_time = last_time - (settings.numSamples+1) * settings.intervalTime;
int64_t old_index = sf.mapTimeToBucketIndex(old_time, true);
assert_equal(-10239, old_index);
assert_equal(-1, sf.mapBucketIndexToFileIndex(old_index));
}
//test firstBucketTime and lastBucketTime
{
setUp();
StatFile sf("created.sf", Stats(), mm);
assert_equal(1000000020, sf.lastBucketTime());
assert_equal(1000000020 - (10240-1)*10, sf.firstBucketTime());
assert_equal(1000000000, sf.bucket(sf.firstBucket()).time());
}
}
bool StatFile::rawUpdateBucket(Bucket const &data, RawUpdateMode mode)
{
if (!fileWritable_)
{
throw std::logic_error(std::string("Attempt to update bucket in read-only StatFile ")
+ fileHeader_->name);
}
// Work out which bucket to put the data in
int64_t targetBucketIndex = mapTimeToBucketIndex(data.time(), false);
int64_t targetBucketTime = data.time() - data.time() % fileHeader_->cfg_interval;
int64_t latestBucketIndex = fileHeader_->last_bucket;
Bucket *bp = 0;
if (targetBucketIndex > latestBucketIndex)
{
// Note: I don't clear the bucket here. It will contain
// old data. The user will have to detect this and discard
// the data instead.
fileHeader_->last_bucket = targetBucketIndex;
fileHeader_->last_time = targetBucketTime;
if (fileHeader_->first_bucket + bucketCount_ <= fileHeader_->last_bucket)
{
fileHeader_->first_bucket = fileHeader_->last_bucket - bucketCount_ + 1;
}
}
if (!StatFile::isBucketIndexInFile(targetBucketIndex))
{
Log(LL_Warning, "libistat") << "Cannot go back to time" << data.time() <<
"when at" << fileHeader_->last_time << ":" << fileHeader_->name;
return false;
}
if (targetBucketIndex < fileHeader_->first_bucket) {
fileHeader_->first_bucket = targetBucketIndex;
}
if (fileHeader_->flags & FILE_FLAG_IS_TRAILING)
{
bp = getTrailingBucket(targetBucketTime);
}
else
{
bp = writableBucket(targetBucketIndex);
if (bp->time() != targetBucketTime)
{
memset(bp, 0, sizeof(*bp));
}
}
if((fileHeader_->flags & FILE_FLAG_IS_COLLATED)
|| mode == RAWUP_OVERWRITE
|| (mode == RAWUP_FILL_EMPTY && bp->count() == 0))
{
memcpy(bp, &data, sizeof(*bp));
}
else
{
Bucket timeData(data, targetBucketTime);
bp->update(timeData);
if(fileHeader_->flags & FILE_FLAG_IS_COUNTER_AGGREGATE)
{
bp->setCount(fileHeader_->fixed_count);
}
}
fileHeader_->cumulative_sum += data.sum();
fileHeader_->cumulative_sum_sq += data.sumSq();
fileHeader_->cumulative_count += data.count();
return true;
}
