This patch insures that TTreeFormula only call GetEntry once per branch
per entry (This speeds up evaluating formula where the same branch is use
several times).

Update TTreeIndex (and hence TTree::BuildIndex) to be able to
work correctly independently of the state of the TTree object (It previously
__required__ TTree::fReadEntry to be equal to -1).

Make TTreeIndex::Print platform independent (it was not
printing correctly on windows).


git-svn-id: http://root.cern.ch/svn/root/trunk@11046 27541ba8-7e3a-0410-8455-c3a389f83636

Fix in TGraphEditor following a recent change in the GUI functions


git-svn-id: http://root.cern.ch/svn/root/trunk@11045 27541ba8-7e3a-0410-8455-c3a389f83636

git-svn-id: http://root.cern.ch/svn/root/trunk@11044 27541ba8-7e3a-0410-8455-c3a389f83636

- The errors boxes attributes (option E2) were not right when the graph
  was painted in an existing TFrame.


git-svn-id: http://root.cern.ch/svn/root/trunk@11043 27541ba8-7e3a-0410-8455-c3a389f83636

Solves the missing break in long lines on windows command prompt.


git-svn-id: http://root.cern.ch/svn/root/trunk@11042 27541ba8-7e3a-0410-8455-c3a389f83636

TStreamerElement::fSize is set to the actual size of the element.  Since it
already takes into account the number of element of the array (if any),
there is no need to remultiply (and this leads to very large number if the
array is large!).  This can be a problem only in the case of emulated
classes.


git-svn-id: http://root.cern.ch/svn/root/trunk@11041 27541ba8-7e3a-0410-8455-c3a389f83636

cases like: /* */ /*. Bug was reported by Pedro Miguel Esteves Ramalhete.


git-svn-id: http://root.cern.ch/svn/root/trunk@11040 27541ba8-7e3a-0410-8455-c3a389f83636

rootcint_tmp.exe.


git-svn-id: http://root.cern.ch/svn/root/trunk@11039 27541ba8-7e3a-0410-8455-c3a389f83636

This patch corrects the dependency of rootcint_tmp for the metautils
dictionary.  It was wrong for windows and was breaking (surprisingly enough
:) ) only the parallel build.


git-svn-id: http://root.cern.ch/svn/root/trunk@11038 27541ba8-7e3a-0410-8455-c3a389f83636

Fix a bug in TMath::Hash reported by Mathieu de Naurois


git-svn-id: http://root.cern.ch/svn/root/trunk@11037 27541ba8-7e3a-0410-8455-c3a389f83636

draw the histogram in a canvas and invoke the editor.


git-svn-id: http://root.cern.ch/svn/root/trunk@11036 27541ba8-7e3a-0410-8455-c3a389f83636

- The protections added in revision 1.16 in LegoFunction must be after the
  logarithms computation (ab[..]) because the limit values stored in
  Hparam (xmin, xmax etc ...) are already logarithmic.


git-svn-id: http://root.cern.ch/svn/root/trunk@11035 27541ba8-7e3a-0410-8455-c3a389f83636

the inspected object.
In the TInspectcanvas destructor do not delete the objects in the list
of objects in the inspector. Deleting objects with the kCanDelete bit set
had side-effects for objects in the stack.


git-svn-id: http://root.cern.ch/svn/root/trunk@11034 27541ba8-7e3a-0410-8455-c3a389f83636

git-svn-id: http://root.cern.ch/svn/root/trunk@11033 27541ba8-7e3a-0410-8455-c3a389f83636

forgot to remove a work-around in mlp, for a problem that was fixed by the
previous patch.


git-svn-id: http://root.cern.ch/svn/root/trunk@11032 27541ba8-7e3a-0410-8455-c3a389f83636

git-svn-id: http://root.cern.ch/svn/root/trunk@11031 27541ba8-7e3a-0410-8455-c3a389f83636

update to the documentation specifiying the actual result of the
TTree::Draw in case of 2 variables.


git-svn-id: http://root.cern.ch/svn/root/trunk@11030 27541ba8-7e3a-0410-8455-c3a389f83636

In TH1::GetAsymmetry the compuation of the bin errors was not correct
From the Jonathan's original mail
"Calculation of histogram errors in TH1::GetAsymmetry appears to be incorrect:

error = 2*TMath::Sqrt(a*a*c2*c2*db*db + c2*c2*b*b*da*da+a*a*b*b*dc2*dc2)/(a+b);

Should read:

error = 2*TMath::Sqrt(a*a*c2*c2*db*db + c2*c2*b*b*da*da +
a*a*b*b*dc2*dc2)/bot/bot;

a) Factor (a+b) in denominator should take into account weight c2. So should be
(a+c2*b) or simply "bot"

b) Factor (a+b) in denominator should also be squared

A = (a - c2*b) / (a + c2*b);
dA/da = 2*c2*b/(a+c2*b)2 ; dA/db = -2*c2*a/(a+c2*b)2; dA/dc2 = -2*a*b/(a+c2*b)2;

Which gives:

error = (2 / (a+c2*b)2 ) * sqrt( c2*c2*b*b*da*da + c2*c2*a*a*db*db +
a*a*b*b*dc2*dc2);


git-svn-id: http://root.cern.ch/svn/root/trunk@11029 27541ba8-7e3a-0410-8455-c3a389f83636

The diffs for the changes from Mihaela. They implement the SavePrimitive method for:
- all shapes
- materials, mixtures and media
- matrices
- TGeoVolume - just a first version

What is missing:
Divisions + manager metod to save:
- all geometry
- just the branch corresponding to one volume

The differences compared to g2root are that:
- when saving just a a branch just materials/media/matrices for that branch are created (module encapsulation)
- instead of creating objects with numerical arguments, the coresponding variables are defined and used.
 e.g. instead of :
   new TGeoBBox("name", 10.,20.,30); the generated code is:

   dx = 10.;
   dy = 20.;
   dz = 30.;
   pShape = new TGeoBBox("name", dx,dy,dz);


The diffs contain also 2 shapes with boundary-safe algorithms: TGeoPara and TGeoTrap


git-svn-id: http://root.cern.ch/svn/root/trunk@11028 27541ba8-7e3a-0410-8455-c3a389f83636

git-svn-id: http://root.cern.ch/svn/root/trunk@11027 27541ba8-7e3a-0410-8455-c3a389f83636

Fix in TCanvas::Streamer to correctly save the colors created
via the color widget and when there is a gap in the list of colors.


git-svn-id: http://root.cern.ch/svn/root/trunk@11026 27541ba8-7e3a-0410-8455-c3a389f83636

I've created a tutorial macro showing the use of an ANN for regression analysis: given a set {i} of input vectors i and a set {o} of output vectors o, one looks for the unknown function f with f(i)=o. The ANN can approximate this function; the DrawTruthDeviation methods can be used to evaluate the quality of the approximation.

Most people know ANNs only as classification tools, not for regression, that's why I though it might be useful.


git-svn-id: http://root.cern.ch/svn/root/trunk@11025 27541ba8-7e3a-0410-8455-c3a389f83636

This patch fixes a bug in DrawNetwork, where the hists' upper edge was smaller than the lower edge (this was causing the corrupted histos in the mlpHiggs tutorial). I updated the new regression methods (some null pointer checks, better labels) and their doc. I added the following comment to the doc of TMultiLayerPerceptron: "(One should still try to pass normalized inputs, e.g. between [0.,1])", and added labels for the output nodes in Draw.


git-svn-id: http://root.cern.ch/svn/root/trunk@11024 27541ba8-7e3a-0410-8455-c3a389f83636

win32gcc needs the x11 inc path, otherwise it'll pick up the wrong headers.


git-svn-id: http://root.cern.ch/svn/root/trunk@11023 27541ba8-7e3a-0410-8455-c3a389f83636

git-svn-id: http://root.cern.ch/svn/root/trunk@11022 27541ba8-7e3a-0410-8455-c3a389f83636

With this patch an embedded editor can be activated via View/Editor menu
in any canvas created in a root session.


git-svn-id: http://root.cern.ch/svn/root/trunk@11021 27541ba8-7e3a-0410-8455-c3a389f83636

patch for a special case in matrix-vector multiplication:
The following scenario does the wrong memory management :

Double_t *a = new Double_t[n];
TVector v; v.Use(n,a)
TMatrixD m(5,5);
.
.
v *= a;

The vector v has the same length after the matrix multiplication, so
in principle there is no need for allocation of new space .
Nevertheless, the previous code did it and failed in the case the
vector
storage was used and not owned .

Now we have the correct behavior:
- if the vector is using another array for storage and the
multiplication
   would mean a resizing, an error message is produced (as before)
- if the vector is using another array for storage and the
multiplication
   keeps the vector length/indexing the same, the multiplication
   can proceed !


git-svn-id: http://root.cern.ch/svn/root/trunk@11020 27541ba8-7e3a-0410-8455-c3a389f83636

  The problem was due to a non-orthodox splitting of a line in a debug
  statement.


git-svn-id: http://root.cern.ch/svn/root/trunk@11019 27541ba8-7e3a-0410-8455-c3a389f83636

Fix in TGenCollectionProxy::PopProxy to correctly initialize fEnv.


git-svn-id: http://root.cern.ch/svn/root/trunk@11018 27541ba8-7e3a-0410-8455-c3a389f83636

git-svn-id: http://root.cern.ch/svn/root/trunk@11017 27541ba8-7e3a-0410-8455-c3a389f83636

git-svn-id: http://root.cern.ch/svn/root/trunk@11016 27541ba8-7e3a-0410-8455-c3a389f83636

the automatic binning from the buffer. These changes affect
the correct running of DrawTest.sh. To be investigated.


git-svn-id: http://root.cern.ch/svn/root/trunk@11015 27541ba8-7e3a-0410-8455-c3a389f83636

- emit signals and change button state on release


git-svn-id: http://root.cern.ch/svn/root/trunk@11014 27541ba8-7e3a-0410-8455-c3a389f83636

sort is ambiguous with a variable in one include on Alpha


git-svn-id: http://root.cern.ch/svn/root/trunk@11013 27541ba8-7e3a-0410-8455-c3a389f83636

is greater or equal to the upper limit.

In TProfile2D constructors force an additional call to SetBuffer.
Because SetBuffer is called from the TH2 constructors, TH1::setBuffer
was called instead of TProfile2D::SetBuffer.

Fix a bug in TProfile2D::BufferEmpty.


git-svn-id: http://root.cern.ch/svn/root/trunk@11012 27541ba8-7e3a-0410-8455-c3a389f83636

fUpdating was set to kTRUE too early in TCanvas::Update in case of Windows
or multiple threads.


git-svn-id: http://root.cern.ch/svn/root/trunk@11011 27541ba8-7e3a-0410-8455-c3a389f83636

git-svn-id: http://root.cern.ch/svn/root/trunk@11010 27541ba8-7e3a-0410-8455-c3a389f83636

which they are called).


git-svn-id: http://root.cern.ch/svn/root/trunk@11009 27541ba8-7e3a-0410-8455-c3a389f83636

make debug output conditional on gDebug.


git-svn-id: http://root.cern.ch/svn/root/trunk@11008 27541ba8-7e3a-0410-8455-c3a389f83636

Fix a problem with shutters.


git-svn-id: http://root.cern.ch/svn/root/trunk@11007 27541ba8-7e3a-0410-8455-c3a389f83636