public void ComputeStats(int index, List <Row> samples)
{
int nNulls = 0;
List <Value> values = new List <Value>();
foreach (var r in samples)
{
Value val = r[index];
if (val is null)
{
nNulls++;
}
values.Add(val);
}
n_distinct_ = values.Distinct().Count();
if (n_distinct_ <= MCVList.NValues_)
{
mcv_ = new MCVList();
var groups = from value in values group value by value into newGroup orderby newGroup.Key select newGroup;
int i = 0;
foreach (var g in groups)
{
mcv_.values_[i] = g.Key;
mcv_.freqs_[i] = (1.0 * g.Count()) / values.Count();
i++;
}
mcv_.nvalues_ = i;
mcv_.validateThis();
}
else
{
// now sort the values and create equal-depth historgram
values.Sort();
int nbuckets = Math.Min(Historgram.NBuckets_, values.Count);
int depth = values.Count / nbuckets;
Debug.Assert(depth >= 1);
hist_ = new Historgram();
for (int i = 0; i < nbuckets; i++)
{
hist_.buckets_[i] = values[(i + 1) * depth - 1];
hist_.distincts_[i] = values.GetRange(i * depth, depth).Distinct().Count();
Debug.Assert(hist_.distincts_[i] > 0);
}
hist_.depth_ = depth;
hist_.nbuckets_ = nbuckets;
}
// finalize the stats
n_rows_ = samples.Count;
Debug.Assert(nNulls <= samples.Count);
if (samples.Count != 0)
{
nullfrac_ = nNulls / samples.Count;
}
}
public void ComputeStats(int index, List <Row> samples)
{
int nNulls = 0;
List <Value> values = new List <Value>();
foreach (var r in samples)
{
Value val = r[index];
if (val is null)
{
nNulls++;
continue;
}
values.Add(val);
}
n_distinct_ = (ulong)values.Distinct().Count();
// initialize mcv whenever the attr is not unique key
if (n_distinct_ < (ulong)values.Count())
{
mcv_ = new MCVList();
var groups = from value in values group value by value into newGroup select newGroup;
Dictionary <Value, int> sortgroup = new Dictionary <Value, int>();
foreach (var g in groups)
{
sortgroup.Add(g.Key, g.Count());
}
// use top 100 values to calculate frequency, ensure that pairs are sorted in a fixed order.
var sorted = from pair in sortgroup orderby pair.Value descending, pair.Key descending select pair;
mcv_.nvalues_ = (int)Math.Min(n_distinct_, MCVList.NValues_);
int i = 0;
double freq = 0.0;
foreach (var g in sorted)
{
mcv_.values_[i] = g.Key;
mcv_.freqs_[i] = (1.0 * g.Value) / values.Count();
freq += mcv_.freqs_[i];
i++;
if (i >= mcv_.nvalues_)
{
break;
}
}
if (n_distinct_ > (ulong)mcv_.nvalues_)
{
Debug.Assert(freq > 0 && freq < 1 + StatConst.epsilon_);
mcv_.otherfreq_ = (1.0 - freq) / (n_distinct_ - (ulong)mcv_.nvalues_);
mcv_.totalfreq_ = freq;
}
else
{
mcv_.otherfreq_ = 0;
mcv_.totalfreq_ = StatConst.one_;
}
mcv_.validateThis();
// remove all values present in mcv
values.RemoveAll(x => mcv_.values_.Contains(x));
}
// initialize histogram unless all values in mcv
if (values.Count > 0)
{
// now sort the values and create equal-depth historgram
values.Sort();
int nbuckets = Math.Min(Historgram.NBuckets_, values.Count);
double depth = ((double)values.Count) / nbuckets;
Debug.Assert(depth >= 1);
hist_ = new Historgram();
for (int i = 0; i < nbuckets; i++)
{
hist_.buckets_[i] = values[Math.Min((int)(i * depth), values.Count - 1)];
}
hist_.depth_ = depth;
hist_.nbuckets_ = nbuckets;
}
// finalize the stats
n_rows_ = (ulong)samples.Count;
Debug.Assert(nNulls <= samples.Count);
if (samples.Count != 0)
{
nullfrac_ = nNulls / samples.Count;
}
}