class TDataSetManager: public TObject

 Main constructor

 Destructor

 Parse the opts string and set the init bits accordingly
 Available options:
    Cq:               set kCheckQuota
    Ar:               set kAllowRegister
    Av:               set kAllowVerify
    Ti:               set kTrustInfo
    Sb:               set kIsSandbox
    Ca:               set kUseCache or kDoNotUseCache
 The opts string may also contain additional unrelated info: in such a case
 the field delimited by the prefix "opt:" is analyzed, e.g. if opts is
 "/tmp/dataset  opt:Cq:-Ar: root://lxb6046.cern.ch" only the substring
 "Cq:-Ar:" will be parsed .

 Read group config file 'cf'.
 If cf == 0 re-read, if changed, the file pointed by fGroupConfigFile .

 expects the following directives:
 Group definition:
   group <groupname> <user>+
 disk quota
   property <groupname> diskquota <quota in GB>
 average filesize (to be used when the file size is not available)
   averagefilesize <average size>{G,g,M,m,K,k}

 Static utility function to gt the number of bytes from a string
 representation in the form "<digit><sfx>" with <sfx> = {"", "k", "M", "G",
 "T", "P"} (case insensitive).
 Returns -1 if the format is wrong.

 Utility function used in various methods for user dataset upload.

 Removes the indicated dataset

 Checks if the indicated dataset exits

 Returns all datasets for the <group> and <user> specified by <uri>.
 If <user> is 0, it returns all datasets for the given <group>.
 If <group> is 0, it returns all datasets.
 The returned TMap contains:
    <group> --> <map of users> --> <map of datasets> --> <dataset> (TFileCollection)

 The unsigned int 'option' is forwarded to GetDataSet and BrowseDataSet.
 Available options (to be .or.ed):
    kShowDefault    a default selection is shown that include the ones from
                    the current user, the ones from the group and the common ones
    kPrint          print the dataset content
    kQuotaUpdate    update quotas
    kExport         use export naming

 NB1: options "kPrint", "kQuoatUpdate" and "kExport" are mutually exclusive
 NB2: for options "kPrint" and "kQuoatUpdate" return is null.

 Scans the dataset indicated by 'uri' following the 'opts' directives

 The 'opts' string contains up to 4 directive fields separated by ':'

  'selection' field :
    A, allfiles:    process all files
    D, staged:      process only staged (on Disk) files (if 'allfiles:' is not specified
                    the default is to process only files marked as non-staged)
  'pre-action field':
    O, open:        open the files marked as staged when processing only files
                    marked as non-staged
    T, touch:       open and touch the files marked as staged when processing
                    only files marked as non-staged
    I, nostagedcheck: do not check the actual stage status on selected files

  'process' field:
    N, noaction:    do nothing on the selected files
    P, fullproc:    open the selected files and extract the meta information
    L, locateonly:  only locate the selected files
    S, stageonly:   issue a stage request for the selected files not yet staged

  'auxilliary' field
    V, verbose:     notify the actions

 Returns 0 on success, -1 if any failure occurs.

 Scans the dataset indicated by <uri> and returns the number of missing files.
 Returns -1 if any failure occurs.
 For more details, see documentation of
 ScanDataSet(TFileCollection *dataset, const char *option)

 Gets quota information from this dataset

 Display quota information

 Prints the quota

 Log info to the monitoring server

 Returns the used space of that group

 returns the quota a group is allowed to have

 updates the used space maps

 Register a dataset, perfoming quota checkings, if needed.
 Returns 0 on success, -1 on failure

 Save into the <datasetdir>/dataset.list file the name of the last updated
 or created or modified dataset
 Returns 0 on success, -1 on error

 Clear cached information matching uri

 Show cached information matching uri

 Creates URI for the dataset manger in the form '[[/dsGroup/]dsUser/]dsName[#dsObjPath]',
 The optional dsObjPath can be in the form [subdir/]objname]'.

 Parses a (relative) URI that describes a DataSet on the cluster.
 The input 'uri' should be in the form '[[/group/]user/]dsname[#[subdir/]objname]',
  where 'objname' is the name of the object (e.g. the tree name) and the 'subdir'
 is the directory in the file wher it should be looked for.
 After resolving against a base URI consisting of proof://masterhost/group/user
 - meaning masterhost, group and user of the current session -
 the path is checked to contain exactly three elements separated by '/':
 group/user/dsname
 If wildcards, '*' is allowed in group and user and dsname is allowed to be empty.
 If onlyCurrent, only group and user of current session are allowed.
 Only non-null parameters are filled by this function.
 Returns kTRUE in case of success.

 Partition dataset 'ds' accordingly to the servers.
 The returned TMap contains:
                <server> --> <subdataset> (TFileCollection)
 where <subdataset> is the subset of 'ds' on <server>
 The partitioning is done using all the URLs in the TFileInfo's, so the
 resulting datasets are not mutually exclusive.
 The string 'exclude' contains a comma-separated list of servers to exclude
 from the map.

 Formatted printout of the content of TFileCollection 'fc'.
 Options in the form
           popt = u * 10 + f
     f    0 => header only, 1 => header + files
   when printing files
     u    0 => print file name only, 1 => print full URL

 Prints formatted information about the dataset 'uri'.
 The type and format of output is driven by 'opt':

   1. opt = "server:srv1[,srv2[,srv3[,...]]]"
            Print info about the subsets of 'uri' on servers srv1, srv2, ...
   2. opt = "servers[:exclude:srv1[,srv2[,srv3[,...]]]]"
            Print info about the subsets of 'uri' on all servers, except
            the ones in the exclude list srv1, srv2, ...
   3. opt = <any>
            Print info about all datasets matching 'uri'

   If 'opt' contains 'full:' the list of files in the datasets are also printed.
   In case 3. this is enabled only if 'uri' matches a single dataset.

   In case 3, if 'opt' contains
      'full:'      the list of files in the datasets are also printed.
      'forcescan:' the dataset are open to get the information; otherwise the
                   pre-processed information is used.
      'noheader:'  the labelling header is not printed; usefull when to chain
                   several printouts
      'noupdate:'  do not update the cache (which may be slow on very remote
                   servers)
      'refresh:'   refresh the information (requires appropriate credentials;
                   typically it can be done only for owned datasets)

 Go through the files in the specified dataset, selecting files according to
 'fopt' and doing on these files the actions described by 'sopt'.
 If required, the information in 'dataset' is updated.

 The int fopt controls which files have to be processed (or added to the list
 if ropt is 1 - see below); 'fopt' is defined in term of csopt and fsopt:
                    fopt = sign(fsopt) * csopt * 100 + fsopt
 where 'fsopt' controls the actual selection
    -1              all files in the dataset
     0              process only files marked as 'non-staged'
   >=1              as 0 but files that are marked 'staged' are open
   >=2              as 1 but files that are marked 'staged' are touched
    10              process only files marked as 'staged'; files marked as 'non-staged'
                    are ignored
 and 'csopt' controls if an actual check on the staged status (via TFileStager) is done
     0              check that the file is staged using TFileStager
     1              do not hard check the staged status
 (example: use fopt = -101 to check the staged status of all the files, or fopt = 110
  to re-check the stage status of all the files marked as staged)

 If 'dbg' is true, some information about the ongoing operations is reguraly
 printed; this can be useful when processing very large datasets, an operation
 which can take a very long time.

 The int 'sopt' controls what is done on the selected files (this is effective only
 if ropt is 0 or 2 - see below):
    -1              no action (fopt = 2 and sopt = -1 touches all staged files)
     0              do the full process: open the files and fill the meta-information
                    in the TFileInfo object, including the end-point URL
     1              only locate the files, by updating the end-point URL (uses TFileStager::Locate
                    which is must faster of an TFile::Open)
     2              issue a stage request on the files

 The int 'ropt' controls which actions are performed:
     0              do the full process: get list of files to process and process them
     1              get the list of files to be scanned and return it in flist
     2              process the files in flist (according to sopt)
 When defined flist is under the responsability the caller.

 If avgsz > 0 it is used for the final update of the dataset global counters.

 If 'mss' is defined use it to initialize the stager (instead of the Url in the
 TFileInfo objects)

 If maxfiles > 0, select for processing a maximum of 'filesmax' files (but if fopt is 1 or 2
 all files marked as 'staged' are still open or touched)

 Return code
     1 dataset was not changed
     2 dataset was changed

 The number of touched, opened and disappeared files are returned in the respective
 variables, if these are defined.

 Open the file described by 'fileinfo' to extract the relevant meta-information.
 Return 0 if OK, -2 if the file cannot be open, -1 if it is corrupted

 Navigate the directory 'd' (and its subdirectories) looking for TTree objects.
 Fill in the relevant metadata information in 'fi'. The name of the TFileInfoMeta
 metadata entry will be "/dir1/dir2/.../tree_name".
 Return 0 on success, -1 if any problem happens (object found in keys cannot be read,
 for example)

 Create a server mapping list from the content of 'srvmaps'
 Return the list (owned by the caller) or 0 if no valid info could be found)

 Static getter for server mapping list

 Check if the dataset server mappings apply to the url defined by 'furl'.
 Use srvmaplist if defined, else use the default list.
 If yes, resolve the mapping into file1 and return kTRUE.
 Otherwise return kFALSE.

 Update scan counters

 for dataset files (init from DataSet.SrvMap)