Software for small angle scattering - Data Formats - Proposal for a simple XML data format - Fortran-callable navigator and extractor for XML data files

Data Formats | Fortran-callable navigator and extractor for XML data files

R. Ghosh, ILL, July 2007

These routines written in C++ use a GPL xml parser written by Frank Vanden Berghen and extend it as a Fortran wrapper to perform basic navigation and extraction of selected data fields in xml files. This enables, for example, for all instances of a data item to be extracted from an xml file without knowing details of the internal structure, merely the actual node name.

Further development will include the ability to start the selection from a chosen branch in the file, allowing the possibility of navigating easily through similarly named segments (nodes) and returning attributes.

The basic routines are:

rxmlopen File opening routine

      numnod=rxmlopen(filename)
      numnod integer return is the number of nodes found
      filename character (in) filename (no trailing blanks)

mgetf Floating point number extractor from named node(s) q.v. igetf - forrtran version

      nfound=mgetf(ff,lim,nodname)
      nfound integer number of reals found
      ff      reals returned
      lim     (in) integer array limit
      nodname (in) character node name (no trailing blanks)

rgetc text field extractor from named node(s)

      nfound = rgetc(buf,iblen,nfield,nth,name) 
      nfound number of characters returned
      buf    characrcter buffer
      iblen  integer buffer length
      nfield integer number of text field in node (starts at 0)
      nth    integer (in) nth occurrence of this node name (starts at 1)
      name   character (in) node name (no trailing blanks)

rnodeinfo node information retrieval (becomes "current_node")

      ifo=rnodeinfo(ntot,nsnod,natt,ntext,nth,nname) 
      ifo    integer <0 no info/node absent
      ntot   integer total number of such named nodes
      nsnod  integer number of sub-nodes
      natt   integer number of attributes
      ntext  integer number of text fields
      nth    integer (in) results for nth node (start at 1)
      nname  character (in) node name

curatts attributes (name,value) of "current_node"

      ifo    integer <0 not found or no "current_node"
      attnam character name
      lenn   integer length of string returned for name
      attval character string value of attribute
      lenv   integer length of string returned for value

The open routine rxmlopen calls Frank Vanden Berghen's xmlParser.cpp. This parses the whole file, finding all nodes. The Fortran callable routine then traverses the parsed names, and builds up a local list of node names which can then scanned efficiently by the extractor routines. This allows all instances of a variable, for example, to be obtained in one pass. The igetf Fortran routine is less efficient since each item is subject to restarting the search. The present version of rxml.cpp will accommodate up to 10000 node names.

A simple test example below shows use to get values from a file

      program frxml
c***** tests reading xml files...
      character ccc*100,fname*50,nname*20,attnam*60,attval*60
      integer rxmlopen,rgetf,rnodeinfo
      real xx(500)
      write(6,1)
1     format(' Give filename ',$)
      read(5,*) fname
      ll=index(fname,' ')-1
      numnam=0
c***** C program MUST reconstruct a zero terminated name
c      the string length passed does not include any termination
c      in memory - and strings are concatenated in Fortran storage.       
      numnam = rxmlopen(fname(1:ll))
      write(6,2) numnam
2     format(i5,' names found')
      lim=500
      write(6,6) 'to find reals (library)  : '
6     format(' Give NAME ',a,$)
      read(5,*) nname
      ll=index(nname,' ')-1
      nfound=mgetf(xx,lim,nname(1:ll))
      write(6,*) 'Found reals ',nfound
      write(6,*) (xx(i),i=1,nfound)
      write(6,6) 'to find reals (fortran)  : '
      read(5,*) nname
      ll=index(nname,' ')-1
      lim=500
      nfound=igetf(xx,lim,nname(1:ll))
      write(6,*) 'Found reals ',nfound
      write(6,*) (xx(i),i=1,nfound)
      lim=500
      write(6,6) 'info for : '
      read(5,*) nname
      nth=1
c***** nth instance 
      ll=index(nname,' ')-1
      ifo=rnodeinfo(ntot,nsnod,natt,ntext,nth,nname(1:ll))
      if(ifo.lt.0) stop 'no info'
      write(6,10)ifo,ntot,nsnod,natt,ntext
10    format(i10,'  ifo'/
     1       i10,'  ntot'/
     1       i10,'  nsnod'/
     1       i10,'  natt'/
     1       i10,'  ntext')
      if(natt.gt.0) then
      do i=1,natt
        ii=i-1
        ifo=myatts(ii,attnam,lenn,attval,lenv)
	if(ifo.ge.0) write(6,*) '   Attribute "',attnam(1:lenn),
     1'"   Value :',attval(1:lenv)
      end do
      end if
      end
	
      integer function igetf(ff,lim,name)
      character *(*) name
      character*1024 buf
      real ff(lim)
      integer iblen,ierr
      integer rnodeinfo,rgetc
      igetf=-1111
      iblen=1024
      limm=lim
c***** only read first field (starts 0)
      nfield=0
      nth=1
     
      ifo=rnodeinfo(ntot,nsnod,natt,ntext,nth,name)
      if(ifo.lt.0) return
      if(ntot.gt.lim) ntot=lim
      ierr=0
      do i=1,ntot
      ff(i)=0.
      nth=i
      nfound = rgetc(buf,iblen,nfield,nth,name) +1
c c-call has two extra string length arguments..
c note differences always with MSVC / gcc
c MSVC has length immediately after string; 
c g77 has lengths added at end 
      if(nfound.le.0) then
        ierr=ierr+1
      else
c        write(6,*) nfound,buf(1:nfound)
        read(buf(1:nfound),*) ff(i)
      end if
      end do
      igetf=ntot
      if(ierr.gt.0) igetf=-ierr
      return
      end

The internal floating point routine or the text routine is used to select specific variables. The first is more efficient since the search through the nodes is only made as a single pass.

RESULTS

> rxml
 Give filename sk1.xml 
  736 names found
 Give NAME to find reals (library)  : I_cm-1
 Found reals  140
  34.3499985  18.9500008  10.8900003  6.78000021  4.52099991  3.19700003
  2.35700011  1.83099997  1.46399999  1.227  1.03499997  0.910099983
  0.796899974  0.721300006  0.66170001  0.605400026  0.57980001  0.555700004
  0.534600019  0.513199985  0.49970001  0.477100015  0.469199985  0.465600014
                              :
                              :
  0.435099989  0.454400003  0.450599998  0.444400012  0.485199988  0.508000016
  0.518999994  0.524500012
 Give NAME to find reals (fortran)  : Q_A-1
 Found reals  140
  0.00700000022  0.00899999961  0.0109999999  0.0130000003  0.0149999997
  0.0170000009  0.0189999994  0.0209999997  0.023  0.0250000004  0.0270000007
  0.0289999992  0.0309999995  0.0329999998  0.0350000001  0.0370000005
  0.0390000008  0.0410000011  0.0430000015  0.0450000018  0.0469999984
                              :
                              :
  0.268999994  0.270999998  0.273000002  0.275000006  0.27700001  0.279000014
  0.280999988  0.282999992  0.284999996
 Give NAME info for : SASprocess
         0  ifo
         1  ntot
        10  nsnod
         1  natt
         0  ntext
    Attribute "name"   Value :COLETTE

A more complete example shows the conversion of an xml file to standard ILL SANS data.

The routines, updates and examples will be posted shortly on the ILL ftp server.