[hist] Fix computation of number of entries in ResetStats

[lmoneta]

When computing the number of entries in TH1::ResetStats always include underflow/overflow. Also in number of FetEffectiveEntries()

Fix a bug in the computation of GetAllSumOfWeights() and improve the function to return (as an optional parameter) the total sum of weight square (including underflow/overflows).

Improve the handling of statistics in TH1::Add methods.
When performing additions of histograms with coefficient > 0 but not equal to 1 automatically compute the sum of weight squares, otherwise the bin errors computations (and the error on the statistics, mean, std) will not be correct.

The bug was happening for dimensions < 2. (see ROOT-21253)

This PR fixes #21253 and the old JIRA item https://its.cern.ch/jira/browse/ROOT-10567?filter=-1

For the JIRA item, in case of Addition/Subtraction of histogram, the nuber of entries is now computed using underflow/overflows. But in case of Additions/Subtraciton with coefficient not equal to 1, the histogram will be weighted and the number of effective entries will be used.

[github-actions]

Test Results

    20 files      20 suites   3d 3h 52m 54s ⏱️
 3 772 tests  3 771 ✅ 0 💤 1 ❌
69 396 runs  69 395 ✅ 0 💤 1 ❌

For more details on these failures, see this check.

Results for commit 34b8c9a.

♻️ This comment has been updated with latest results.

[dpiparo]

Thanks @lmoneta, could tests be added for the two issues fixed?

[ferdymercury]

@lmoneta thanks for the fix! For the test, could you go to

https://github.com/lmoneta/root/blob/hist_fix_entries_after_reset/hist/hist/test/test_TH1.cxx#L115

and below all lines starting with

EXPECT_FLOAT_EQ(h*.GetSumOfWeights(), *);

add two additional lines:

EXPECT_FLOAT_EQ(h*.GetSumOfAllWeights(true), *);
EXPECT_FLOAT_EQ(h*.GetSumOfAllWeights(false), *);

[ferdymercury]

You can also copy-paste a similar example into https://github.com/lmoneta/root/blob/hist_fix_entries_after_reset/hist/hist/test/test_TH2.cxx and https://github.com/lmoneta/root/blob/hist_fix_entries_after_reset/hist/hist/test/test_TH3.cxx

TEST(TH2, GetSumOfAllWeights)
{
   TH2F h2("h2", "h2", 10, 0, 1, 10, 0, 1);
   h2.SetBinContent(1, 2, 1.);
   h2.SetBinContent(3, 4, 5.);
   h2.SetBinContent(5, 1, 10.);
   h2.SetBinContent(7, 5, -1.);
   h2.SetBinContent(10, 2, 3.);
   h2.SetBinContent(150, 2, 3.);
   EXPECT_FLOAT_EQ(h2.GetEntries(), 6.);
   EXPECT_FLOAT_EQ(h2.GetSumOfWeights(), 18.);
   EXPECT_FLOAT_EQ(h2.GetSumOfAllWeights(true), 21.);
   EXPECT_FLOAT_EQ(h2.GetSumOfAllWeights(false), 18.);
   double sumsq;
   EXPECT_FLOAT_EQ(h2.GetSumOfAllWeights(true,&sumsq), 21.);
   EXPECT_FLOAT_EQ(sumsq, 145);
}

TEST(TH3, GetSumOfAllWeights)
{
   TH3F h3("h3", "h3", 10, 0, 1, 10, 0, 1, 3, 0, 1);
   h3.SetBinContent(1, 2, 1, 1.);
   h3.SetBinContent(3, 4, 2, 5.);
   h3.SetBinContent(5, 1, 3, 10.);
   h3.SetBinContent(7, 5, 1, -1.);
   h3.SetBinContent(10, 2, 2, 3.);
   h3.SetBinContent(150, 2, 3, 3.);
   EXPECT_FLOAT_EQ(h3.GetEntries(), 6.);
   EXPECT_FLOAT_EQ(h3.GetSumOfWeights(), 18.);
   EXPECT_FLOAT_EQ(h3.GetSumOfAllWeights(true), 21.);
   EXPECT_FLOAT_EQ(h3.GetSumOfAllWeights(false), 18.);
   double sumsq;
   EXPECT_FLOAT_EQ(h3.GetSumOfAllWeights(true,&sumsq), 21.);
   EXPECT_FLOAT_EQ(sumsq, 145);
}

For the JIRA issue, sth like this:

TEST(TH1, Add)
{
   TH1F h1("h1", "h1", 10, 0, 1);
   h1.SetBinContent(1, 1.);
   h1.SetBinContent(3, 5.);
   h1.SetBinContent(5, 10.);
   h1.SetBinContent(7, -1.);
   h1.SetBinContent(10, 3.);
   TH1F h2("h2", "h2", 10, 0, 1);
   h2.SetBinContent(1, 2.);
   h2.SetBinContent(3, 10.);
   h2.SetBinContent(5, 20.);
   h2.SetBinContent(7, -2.);
   h2.SetBinContent(10, 6.);
   h2.Add(&h1, 1.);
   EXPECT_FLOAT_EQ(h2.GetEntries(), 10.);
   EXPECT_FLOAT_EQ(h2.GetSumOfWeights(), ...);
}

[ferdymercury]

Suggested change

	// Also Create Sumw2 if is not a simple addition , otherwise errors will not be correctly computed
	// Also Create Sumw2 if is not a simple addition, otherwise errors will not be correctly computed