class TFile: public TDirectoryFile

 File default Constructor.

 Opens or creates a local ROOT file whose name is fname1. It is
 recommended to specify fname1 as "<file>.root". The suffix ".root"
 will be used by object browsers to automatically identify the file as
 a ROOT file. If the constructor fails in any way IsZombie() will
 return true. Use IsOpen() to check if the file is (still) open.

 To open non-local files use the static TFile::Open() method, that
 will take care of opening the files using the correct remote file
 access plugin.

 If option = NEW or CREATE   create a new file and open it for writing,
                             if the file already exists the file is
                             not opened.
           = RECREATE        create a new file, if the file already
                             exists it will be overwritten.
           = UPDATE          open an existing file for writing.
                             if no file exists, it is created.
           = READ            open an existing file for reading (default).
           = NET             used by derived remote file access
                             classes, not a user callable option
           = WEB             used by derived remote http access
                             class, not a user callable option
 If option = "" (default), READ is assumed.

 The file can be specified as a URL of the form:
    file:///user/rdm/bla.root or file:/user/rdm/bla.root

 The file can also be a member of an archive, in which case it is
 specified as:
    multi.zip#file.root or multi.zip#0
 which will open file.root which is a member of the file multi.zip
 archive or member 1 from the archive. For more on archive file
 support see the TArchiveFile class.

 TFile and its remote access plugins can also be used to open any
 file, i.e. also non ROOT files, using:
    file.tar?filetype=raw
 This is convenient because the many remote file access plugins allow
 easy access to/from the many different mass storage systems.

 The title of the file (ftitle) will be shown by the ROOT browsers.

 A ROOT file (like a Unix file system) may contain objects and
 directories. There are no restrictions for the number of levels
 of directories.

 A ROOT file is designed such that one can write in the file in pure
 sequential mode (case of BATCH jobs). In this case, the file may be
 read sequentially again without using the file index written
 at the end of the file. In case of a job crash, all the information
 on the file is therefore protected.

 A ROOT file can be used interactively. In this case, one has the
 possibility to delete existing objects and add new ones.
 When an object is deleted from the file, the freed space is added
 into the FREE linked list (fFree). The FREE list consists of a chain
 of consecutive free segments on the file. At the same time, the first
 4 bytes of the freed record on the file are overwritten by GAPSIZE
 where GAPSIZE = -(Number of bytes occupied by the record).

 Option compress is used to specify the compression level and algorithm:
  compress = 100 * algorithm + level
  level = 0, objects written to this file will not be compressed.
  level = 1, minimal compression level but fast.

  level = 9, maximal compression level but slower and might use more memory.
 (For the currently supported algorithms, the maximum level is 9)
 If compress is negative it indicates the compression level is not set yet.

 The enumeration ROOT::ECompressionAlgorithm associates each
 algorithm with a number. There is a utility function to help
 to set the value of compress. For example,
   ROOT::CompressionSettings(ROOT::kLZMA, 1)
 will build an integer which will set the compression to use
 the LZMA algorithm and compression level 1.  These are defined
 in the header file Compression.h.

 Note that the compression settings may be changed at any time.
 The new compression settings will only apply to branches created
 or attached after the setting is changed and other objects written
 after the setting is changed.

 In case the file does not exist or is not a valid ROOT file,
 it is made a Zombie. One can detect this situation with a code like:
    TFile f("file.root");
    if (f.IsZombie()) {
       cout << "Error opening file" << endl;
       exit(-1);
    }

  When opening the file, the system checks the validity of this directory.
  If something wrong is detected, an automatic Recovery is performed. In
  this case, the file is scanned sequentially reading all logical blocks
  and attempting to rebuild a correct directory (see TFile::Recover).
  One can disable the automatic recovery procedure when reading one
  or more files by setting the environment variable "TFile::Recover 0"
  in the system.rootrc file.

 TFile objects can not be copied.

 File destructor.

 Initialize a TFile object.
 TFile implementations providing asynchronous open functionality need to
 override this method to run the appropriate checks before calling this
 standard initialization part. See TXNetFile::Init for an example.

 Close a file.
 If option == "R", all TProcessIDs referenced by this file are deleted.
 Calling TFile::Close("R") might be necessary in case one reads a long list
 of files having TRef, writing some of the referenced objects or TRef
 to a new file. If the TRef or referenced objects of the file being closed
 will not be referenced again, it is possible to minimize the size
 of the TProcessID data structures in memory by forcing a delete of
 the unused TProcessID.

 Creates key for object and converts data to buffer.

 Creates key for object and converts data to buffer.

 Return the current ROOT file if any.
 Note that if 'cd' has been called on a TDirectory that does not belong to a file,
 gFile will be unchanged and still points to the file of the previous current
 directory that was a file.

 Delete object namecycle.
 Namecycle identifies an object in the top directory of the file
   namecycle has the format name;cycle
   name  = * means all
   cycle = * means all cycles (memory and keys)
   cycle = "" or cycle = 9999 ==> apply to a memory object
   When name=* use T* to delete subdirectories also

 Examples:
     foo   : delete object named foo in memory
     foo;1 : delete cycle 1 of foo on file
     foo;* : delete all cycles of foo on disk and also from memory
     *;2   : delete all objects on file having the cycle 2
     *;*   : delete all objects from memory and file
    T*;*   : delete all objects from memory and file and all subdirectories

 Fill Graphics Structure and Paint.
 Loop on all objects (memory or file) and all subdirectories.

 Draw map of objects in this file.

 Synchronize a file's in-core and on-disk states.

 Flush the write cache if active.
 Return kTRUE in case of error

 Encode file output buffer.
 The file output buffer contains only the FREE data record.

 Return the best buffer size of objects on this file.
 The best buffer size is estimated based on the current mean value
 and standard deviation of all objects written so far to this file.
 Returns mean value + one standard deviation.

 Return the file compression factor.
 Add total number of compressed/uncompressed bytes for each key.
 return ratio of the two.

 Method returning errno. Is overriden in TRFIOFile.

 Method resetting the errno. Is overridden in TRFIOFile.

 Return a pointer to the current read cache.

 Return a pointer to the current write cache.

 Read the logical record header starting at position first.
 Maxbytes bytes are read into buf the function reads nread bytes
 where nread is the minimum of maxbytes and the number of bytes
 before the end of file. The function returns nread.
 In output arguments:
    nbytes : number of bytes in record
             if negative, this is a deleted record
             if 0, cannot read record, wrong value of argument first
    objlen : uncompressed object size
    keylen : length of logical record header
 Note that the arguments objlen and keylen are returned only
 if maxbytes >=16

 Returns the current file size. Returns -1 in case the file could not
 be stat'ed.

 Returns the cached list of StreamerInfos used in this file.

 Read the list of TStreamerInfo objects written to this file.
 The function returns a TList. It is the user'responsability
 to delete the list created by this function.

 Using the list, one can access additional information,eg:
   TFile f("myfile.root");
   TList *list = f.GetStreamerInfoList();
   TStreamerInfo *info = (TStreamerInfo*)list->FindObject("MyClass");
   Int_t classversionid = info->GetClassVersion();
   delete list;

 List File contents.
 Indentation is used to identify the file tree.
 Subdirectories are listed first, then objects in memory,
 then objects on the file.

 Returns kTRUE in case file is open and kFALSE if file is not open.

 Mark unused bytes on the file.
 The list of free segments is in the fFree linked list.
 When an object is deleted from the file, the freed space is added
 into the FREE linked list (fFree). The FREE list consists of a chain
 of consecutive free segments on the file. At the same time, the first
 4 bytes of the freed record on the file are overwritten by GAPSIZE
 where GAPSIZE = -(Number of bytes occupied by the record).

 List the contents of a file sequentially.
 For each logical record found, it prints:
  Date/Time  Record_Adress Logical_Record_Length  ClassName  CompressionFactor

  Example of output
  20010404/150437  At:64        N=150       TFile
  20010404/150440  At:214       N=28326     TBasket        CX =  1.13
  20010404/150440  At:28540     N=29616     TBasket        CX =  1.08
  20010404/150440  At:58156     N=29640     TBasket        CX =  1.08
  20010404/150440  At:87796     N=29076     TBasket        CX =  1.10
  20010404/150440  At:116872    N=10151     TBasket        CX =  3.15
  20010404/150441  At:127023    N=28341     TBasket        CX =  1.13
  20010404/150441  At:155364    N=29594     TBasket        CX =  1.08
  20010404/150441  At:184958    N=29616     TBasket        CX =  1.08
  20010404/150441  At:214574    N=29075     TBasket        CX =  1.10
  20010404/150441  At:243649    N=9583      TBasket        CX =  3.34
  20010404/150442  At:253232    N=28324     TBasket        CX =  1.13
  20010404/150442  At:281556    N=29641     TBasket        CX =  1.08
  20010404/150442  At:311197    N=29633     TBasket        CX =  1.08
  20010404/150442  At:340830    N=29091     TBasket        CX =  1.10
  20010404/150442  At:369921    N=10341     TBasket        CX =  3.09
  20010404/150442  At:380262    N=509       TH1F           CX =  1.93
  20010404/150442  At:380771    N=1769      TH2F           CX =  4.32
  20010404/150442  At:382540    N=1849      TProfile       CX =  1.65
  20010404/150442  At:384389    N=18434     TNtuple        CX =  4.51
  20010404/150442  At:402823    N=307       KeysList
  20010404/150443  At:403130    N=4548      StreamerInfo   CX =  3.65
  20010404/150443  At:407678    N=86        FreeSegments
  20010404/150443  At:407764    N=1         END

 Paint all objects in the file.

 Print all objects in the file.

 Read a buffer from the file at the offset 'pos' in the file.
 Returns kTRUE in case of failure.
 Compared to ReadBuffer(char*, Int_t), this routine does _not_
 change the cursor on the physical file representation (fD)
 if the data is in this TFile's cache.

 Read a buffer from the file. This is the basic low level read operation.
 Returns kTRUE in case of failure.

 Read the nbuf blocks described in arrays pos and len,
 where pos[i] is the seek position of block i of length len[i].
 Note that for nbuf=1, this call is equivalent to TFile::ReafBuffer.
 This function is overloaded by TNetFile, TWebFile, etc.
 Returns kTRUE in case of failure.

 Read buffer via cache. Returns 0 if the requested block is
 not in the cache, 1 in case read via cache was successful,
 2 in case read via cache failed.

 Read the FREE linked list.
 Every file has a linked list (fFree) of free segments.
 This linked list has been written on the file via WriteFree
 as a single data record.

The TProcessID with number pidf is read from this file.
If the object is not already entered in the gROOT list, it is added.

 Attempt to recover file if not correctly closed.
 The function returns the number of keys that have been recovered.
 If no keys can be recovered, the file will be declared Zombie by
 the calling function. This function is automatically called when
 opening a file.

 If the file is open in read only mode, the file is not modified.
 If open in update mode and the function finds something to recover,
  a new directory header is written to the file. When opening the file gain
  no message from Recover will be reported.
 If keys have been recovered, the file is usable and you can safely
 read the corresponding objects.
 If the file is not usable (a zombie), you can test for this case
 with code like:
   TFile f("myfile.root");
   if (f.IsZombie()) {file is unusable)
 If the file has been recovered, the bit kRecovered is set in the TFile object in memory.
 You can test if the file has been recovered with
   if (f.TestBit(TFile::kRecovered)) {.. the file has been recovered}

 When writing TTrees to a file, it is important to save the Tree header
 at regular intervals (see TTree::AutoSave). If a file containing a Tree
 is recovered, the last Tree header written to the file will be used.
 In this case all the entries in all the branches written before writing
 the header are valid entries.

 One can disable the automatic recovery procedure by setting
 TFile.Recover 0
 in the system.rootrc file.

 Reopen a file with a different access mode, like from READ to
 UPDATE or from NEW, CREATE, RECREATE, UPDATE to READ. Thus the
 mode argument can be either "READ" or "UPDATE". The method returns
 0 in case the mode was successfully modified, 1 in case the mode
 did not change (was already as requested or wrong input arguments)
 and -1 in case of failure, in which case the file cannot be used
 anymore. The current directory (gFile) is changed to this file.

 Set position from where to start reading.

 Seek to a specific position in the file. Pos it either kBeg, kCur or kEnd.

 See comments for function SetCompressionSettings

 See comments for function SetCompressionSettings

 Used to specify the compression level and algorithm:
  settings = 100 * algorithm + level

  level = 0, objects written to this file will not be compressed.
  level = 1, minimal compression level but fast.

  level = 9, maximal compression level but slower and might use more memory.
 (For the currently supported algorithms, the maximum level is 9)
 If compress is negative it indicates the compression level is not set yet.

 The enumeration ROOT::ECompressionAlgorithm associates each
 algorithm with a number. There is a utility function to help
 to set the value of the argument. For example,
   ROOT::CompressionSettings(ROOT::kLZMA, 1)
 will build an integer which will set the compression to use
 the LZMA algorithm and compression level 1.  These are defined
 in the header file Compression.h.

 Note that the compression settings may be changed at any time.
 The new compression settings will only apply to branches created
 or attached after the setting is changed and other objects written
 after the setting is changed.

 Set a pointer to the read cache.
 NOTE:  This relinquish ownership of the previous cache, so if you do not
 already have a pointer to the previous cache (and there was a previous
 cache), you ought to retrieve (and delete it if needed) using:
    TFileCacheRead *older = myfile->GetCacheRead();

 Set a pointer to the write cache.
 If file is null the existing write cache is deleted.

 Return the size in bytes of the file header.

 Stream a TFile object.

 Increment statistics for buffer sizes of objects in this file.

 Dummy function kept for backward compatibility.
 The read  cache is now managed by TFileCacheRead
 The write cache is now managed by TFileCacheWrite
 Both caches are created automatically by the system.

 Write memory objects to this file.
 Loop on all objects in memory (including subdirectories).
 A new key is created in the KEYS linked list for each object.
 The list of keys is then saved on the file (via WriteKeys)
 as a single data record.
 For values of opt see TObject::Write().
 The directory header info is rewritten on the directory header record.
 The linked list of FREE segments is written.
 The file header is written (bytes 1->fBEGIN).

 One can not save a const TDirectory object.

 Write a buffer to the file. This is the basic low level write operation.
 Returns kTRUE in case of failure.

 Write buffer via cache. Returns 0 if cache is not active, 1 in case
 write via cache was successful, 2 in case write via cache failed.

 Write FREE linked list on the file.
 The linked list of FREE segments (fFree) is written as a single data
 record.

 Write File Header.

 Generate code in directory dirname for all classes specified in
 argument classes If classes = "*" (default and currently the
 only supported value), the function generates an include file
 for each class in the StreamerInfo list for which a TClass
 object does not exist.

 The code generated includes:
    dirnameProjectHeaders.h  // contains one #include statement per generated header file
    dirnameProjectSource.cxx // contains all the constructors and destructors implementation.
 and one header per class that is not nested inside another class.
 The header file name is the fully qualified name of the class after all the special characters
 "<>,:" are replaced by underscored.  For example for pair<edm::Vertex,int> the file name is
 pair_edm__Vertex_int_.h

 In the generated classes, map, multimap when the first template parameter is a class
 are replaced by a vector of pair. set and multiset when the tempalte parameter
 is a class are replaced by a vector. This is required since we do not have the
 code needed to order and/or compare the object of the classes.

 If option = "new" (default) a new directory dirname is created.
                   If dirname already exist, an error message is printed
                   and the function returns.
 If option = "recreate", then;
                   if dirname does not exist, it is created (like in "new")
                   if dirname already exist, all existing files in dirname
                   are deleted before creating the new files.
 If option = "update", then new classes are added to the existing directory.
                   Existing classes with the same name are replaced by the
                   new definition. If the directory dirname doest not exist,
                   same effect as "new".
 If option = "genreflex", then use genreflex rather than rootcint to generate
                   the dictionary.
 If option = "par", create a PAR file with the minimal set of code needed to read the content
                   of the ROOT file. The name of the PAR file is basename(dirname), with extension
                   '.par' enforced; the PAR file will be created at dirname(dirname) .
 If, in addition to one of the 3 above options, the option "+" is specified,
 the function will generate:
   - a script called MAKEP to build the shared lib
   - a dirnameLinkDef.h file
   - rootcint will be run to generate a dirnameProjectDict.cxx file
   - dirnameProjectDict.cxx will be compiled with the current options in compiledata.h
   - a shared lib dirname.so will be created.
 If the option "++" is specified, the generated shared lib is dynamically
 linked with the current executable module.
 If the option "+" and "nocompile" are specified, the utility files are generated
 as in the option "+" but they are not executed.
 Example:
  file.MakeProject("demo","*","recreate++");
  - creates a new directory demo unless it already exist
  - clear the previous directory content
  - generate the xxx.h files for all classes xxx found in this file
    and not yet known to the CINT dictionary.
  - creates the build script MAKEP
  - creates a LinkDef.h file
  - runs rootcint generating demoProjectDict.cxx
  - compiles demoProjectDict.cxx into demoProjectDict.o
  - generates a shared lib demo.so
  - dynamically links the shared lib demo.so to the executable
  If only the option "+" had been specified, one can still link the
  shared lib to the current executable module with:
     gSystem->load("demo/demo.so");

 The following feature is not yet enabled:
 One can restrict the list of classes to be generated by using expressions like:
   classes = "Ali*" generate code only for classes starting with Ali
   classes = "myClass" generate code for class MyClass only.

 Create makefile at 'filemake' for PAR package 'pack'.
 Called by MakeProject when option 'par' is given.
 Return 0 on success, -1 on error.

 Create BUILD.sh and SETUP.C under 'proofinf' for PAR package 'pack'.
 Called by MakeProject when option 'par' is given.
 Return 0 on success, -1 on error.

 Read the list of StreamerInfo from this file.
 The key with name holding the list of TStreamerInfo objects is read.
 The corresponding TClass objects are updated.
 Note that this function is not called if the static member fgReadInfo is falsse.
  (see TFile::SetReadStreamerInfo)

 static function to set fgReadInfo.
 If fgReadInfo is true (default) TFile::ReadStreamerInfo is called
  when opening the file.
 It may be interesting to set fgReadInfo to false to speedup the file
 opening time or in case libraries containing classes referenced
 by the file have not yet been loaded.
 if fgReadInfo is false, one can still read the StreamerInfo with
    myfile.ReadStreamerInfo();

 Show the StreamerInfo of all classes written to this file.

 Check if the ProcessID pidd is already in the file,
 if not, add it and return the index  number in the local file list.

 Write the list of TStreamerInfo as a single object in this file
 The class Streamer description for all classes written to this file
 is saved. See class TStreamerInfo.

 Static member function allowing to open a file for reading through the file
 cache. The file will be downloaded to the cache and opened from there.
 If the download fails, it will be opened remotely.
 The file will be downloaded to the directory specified by SetCacheFileDir().

 Static member function allowing the creation/opening of either a
 TFile, TNetFile, TWebFile or any TFile derived class for which an
 plugin library handler has been registered with the plugin manager
 (for the plugin manager see the TPluginManager class). The returned
 type of TFile depends on the file name specified by 'url'.
 If 'url' is a '|'-separated list of file URLs, the 'URLs' are tried
 sequentially in the specified order until a successful open.
 If the file starts with "root:", "roots:" or "rootk:" a TNetFile object
 will be returned, with "http:" a TWebFile, with "file:" a local TFile,
 etc. (see the list of TFile plugin handlers in $ROOTSYS/etc/system.rootrc
 for regular expressions that will be checked) and as last a local file will
 be tried.
 Before opening a file via TNetFile a check is made to see if the URL
 specifies a local file. If that is the case the file will be opened
 via a normal TFile. To force the opening of a local file via a
 TNetFile use either TNetFile directly or specify as host "localhost".
 The netopt argument is only used by TNetFile. For the meaning of the
 options and other arguments see the constructors of the individual
 file classes. In case of error returns 0.

 For TFile implementations supporting asynchronous file open, see
 TFile::AsyncOpen(...), it is possible to request a timeout with the
 option:
  TIMEOUT=<secs>   the timeout must be specified in seconds and
                   it will be internally checked with granularity of
                   one millisec.

 For remote files there is the option:
  CACHEREAD     opens an existing file for reading through the file cache.
                The file will be downloaded to the cache and opened from there.
                If the download fails, it will be opened remotely.
                The file will be downloaded to the directory specified by
                SetCacheFileDir().

 Static member function to submit an open request. The request will be
 processed asynchronously. See TFile::Open(const char *, ...) for an
 explanation of the arguments. A handler is returned which is to be passed
 to TFile::Open(TFileOpenHandle *) to get the real TFile instance once
 the file is open.
 This call never blocks and it is provided to allow parallel submission
 of file opening operations expected to take a long time.
 TFile::Open(TFileOpenHandle *) may block if the file is not yet ready.
 The sequence
    TFile::Open(TFile::AsyncOpen(const char *, ...))
 is equivalent to
    TFile::Open(const char *, ...) .
 To be effective, the underlying TFile implementation must be able to
 support asynchronous open functionality. Currently, only TXNetFile
 supports it. If the functionality is not implemented, this call acts
 transparently by returning an handle with the arguments for the
 standard synchronous open run by TFile::Open(TFileOpenHandle *).
 The retuned handle will be adopted by TFile after opening completion
 in TFile::Open(TFileOpenHandle *); if opening is not finalized the
 handle must be deleted by the caller.

 Waits for the completion of an asynchronous open request.
 Returns the associated TFile, transferring ownership of the
 handle to the TFile instance.

 Interface to system open. All arguments like in POSIX open().

 Interface to system close. All arguments like in POSIX close().

 Interface to system read. All arguments like in POSIX read().

 Interface to system write. All arguments like in POSIX write().

 Interface to system lseek. All arguments like in POSIX lseek()
 except that the offset and return value are of a type which are
 able to handle 64 bit file systems.

 Return file stat information. The interface and return value is
 identical to TSystem::GetPathInfo(). The function returns 0 in
 case of success and 1 if the file could not be stat'ed.

 Interface to system fsync. All arguments like in POSIX fsync().

 Return the total number of bytes written so far to the file.

 Static function returning the total number of bytes read from all files.

 Static function returning the total number of bytes written to all files.
 Does not take into account what might still be in the write caches.

 Static function returning the total number of read calls from all files.

 Static function returning the readahead buffer size.

 Sets the directory where to locally stage/cache remote files.
 If the directory is not writable by us return kFALSE.

 Get the directory where to locally stage/cache remote files.

 We try to shrink the cache to the desired size.
 With the clenupinterval you can specify the minimum amount of time after
 the previous cleanup before the cleanup operation is repeated in
 the cache directory

 Sets open timeout time (in ms). Returns previous timeout value.

 Returns open timeout (in ms).

 Sets only staged flag. Returns previous value of flag.
 When true we check before opening the file if it is staged, if not,
 the open fails.

 Returns staged only flag.

 Return kTRUE if 'url' matches the coordinates of this file.
 The check is implementation dependent and may need to be overload
 by each TFile implememtation relying on this check.
 The default implementation checks the file name only.

 Get final URL for file being opened asynchronously.
 Returns 0 is the information is not yet available.

 Print file copy progress.

 Allows to copy this file to the dst URL. Returns kTRUE in case of success,
 kFALSE otherwise.

 Allows to copy file from src to dst URL. Returns kTRUE in case of success,
 kFALSE otherwise.

 Read specified byte range asynchronously. Actually we tell the kernel
 which blocks we are going to read so it can start loading these blocks
 in the buffer cache.

 Max number of bytes to prefetch. By default this is 75% of the
 read cache size. But specific TFile implementations may need to change it

 Interface for text-based TDirectory I/O