Hi Christian,
You can use TSystem::AccessPathName to test if a file exists and do, eg:
if (gSystem->AccessPathName(fileName)) f = new TFile(fileName,"update")
else f = new TFile(fileName,"create")
The Tree branch creator needs the type information about your classes
in most cases, in particular if your class derives from a compiled class
or if your class has a TClonesArray. You have to compile your classes
for the time being.
Rene Brun
cstrato@EUnet.at wrote:
>
> Dear Rooters
>
> I apologize for sending the following questions again but it seems that
> my message did not arrive at roottalk
>
> Best regards
> Christian Stratowa
>
> Subject:
> root file and custom classes
> Date:
> Sun, 02 Jul 2000 20:00:07 +0100
> From:
> cstrato@EUnet.at
> To:
> roottalk <roottalk@pcroot.cern.ch>
>
> Dear Rooters:
>
> Maybe you could answer the following questions regarding root files:
>
> 1, I have a function ?save_data(file.root, data.txt)?, which creates
> a rootfile and stores the data in a tree. When file.root does already
> exist, the function should update the file and write new data into
> a new tree of the same file. The skeleton of the function looks like
> this:
>
> void save_data(fileName, data)
> {
> TFile *f = new TFile(fileName, ?CREATE?);
> if (!f->Open()) {
> f->Open(fileName, ?UPDATE?;
> //alternatively:
> // delete f;
> // f = new TFile(fileName, ?UPDATE);
> }
>
> // call: MyClassA->SaveData(fileName, data)
>
> f->Write();
> f->Close();
> }
>
> When fileName does not exist, everything is fine, however, when
> I call the function again with different data I get:
> ?Error in <TFile::TFile>: file fileName already exists?
> However the execution continues and f->Open(..) results in:
> ?Warning in <TFile::Write>: file not opened in write mode?
> Interestingly, the data are stored in a new tree as supposed to.
> Why do I get then this warning?
>
> I can avoid the warning by deleting the file first and creating
> a new file f = new TFile(fileName, ?UPDATE) as shown above.
> However, this is not very elegant, so my question is:
> What is the best way to solve this problem?
>
> 2, As seen above, the tree is not created in the function but in
> a new class MyClassA. Furthermore, the data are described in a
> class MyClassB: branchB = tree->Branch(name,?MyClassB?, &myClassB)
> The tree is called from MyClassA and should also store MyClassA:
> branchA = tree->Branch(name,?MyClassA?, &this)
> Running root results in the errors that MyClassA and MyClassB are not
> found.
> When I include ClassDef() and ClassImp() then I get the following
> errors:
> Error: MyClass() header declared but not defined FILE:...
> (class MyClass*) 0x...
> !!!Dictionary position not recovered because G__unloadfile() in macro!!!
>
> Does this mean that I have to compile the classes first to create a
> dictionary?
>
> If I consider the root test example ?MainEvent.cxx?:
> Is there no way to run this example using only the cint interpreter,
> i.e.
> without first compiling Event.h into a shared library to create a
> dictionary?
>
> This would mean that I could not test my own classes interactively, but
> have to compile them everytime I find a bug or want to change/add a
> method.
> Is there a way to avoid this?
>
> Thank you very much in advance for your help.
>
> Best regards
> Christian Stratowa, Vienna, Austria
This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:29 MET