claw43to50.txt

begin_html [use: doc/doc.css]
<!--
     For a more readable version of this file, execute
                  unix>  make htmls
     in this directory and then point your browser to the resulting .html
-->

<h1> Hints on converting from Clawpack 4.3 to Clawpack 5.0 </h1>

<h4> Changes from f77 to f95 </h4>

<ul>
<li> For Clawpack 5.0 a Fortran 95 compiler such as gfortran must be used.
We now use dynamic memory allocation and other newer features.  Files that
use f95 explicitly have the extension .f95.

<li> Most old f77 files ending in .f still work when compiled with gfortran
and can be used together with .f95 files.  So old Riemann solvers or qinit.f
or setprob.f files can generally still be used.  In some of the examples
in claw/examples these have been rewritten as .f95 files to take advantage
of newer features.  If you don't know much Fortran, these are probably
easier to modify without introducing errors.

<li> In Clawpack 4.3 each example directory had a file driver.f to dimension
various arrays (q, work, and aux if needed).  In Clawpack 5.0 this has
disappeared.  Instead the library routine main.f95 (e.g.
[clawcode:clawpack/1d/lib/main.f95] in 1d) uses dynamic memory allocation to
allocate these arrays based on the value of mx and other parameters in the
data files.

<li> In Clawpack 4.3 the main program in driver.f called claw1ez,
which read in data from claw1ez.data and then called claw1 (in 1d, for
example).  In Clawpack 5.0 the main program in main.f95 calls claw1.f95
which reads in data from claw.data.  This probably does not matter to you
unless you were doing something sophisticated with your own version of one
of these routines.

</ul> 

<h4>Changes to the data files</h4>

<ul>
<li> The old data file clawNez.data (N=1,2,3 for the number of space dimensions)
has been replaced by a file that is simply called claw.data in all dimensions.
At the moment this file can be identical to the old clawNez.data with one
exception:
<p>
One new line should be added to the start of this file containing the
integer 1, 2, or 3 for the number of space dimensions.
<li> Note that as far a Fortran is concerned, everything on a line beyond
the number(s) it is trying to read is ignored and is simply a comment.  The
comments have been changed in many of the claw.data files but the set of
parameters read in and their order is the same with the above exception.

<li>  The syntax  <b>value  =:  name</b>  used in claw.data is used by some
Python scripts but ignored by Fortran.   

<li> See the Clawpack 4.3 documentation [link:claw43user.pdf (pdf)] for
more information on all the parameters in claw.data and in amr/amr2ez.data
files used with AMR.

<li> We are considering completely reformulating the format of claw.data and
the manner in which this data is read in by Fortran and Python. See
[claw:dev/namelist_examples].

</ul>

[name:plotting]
<h4>Plotting</h4>

<ul>
<li> Clawpack 5.0 produces the same ascii files fort.q000N and fort.t000N at
output time N as did Clawpack 4.3 with one exception:

<p> 
fort.t000N has a new fourth line containing ndim, the number of space dimensions,
before the fifth line with maux (number of auxiliary variables).

<li> If you want to use the old Matlab scripts from [claw:matlab] these
should still work fine (unless you were using the value maux for some reason,
which the standard scripts do not).  

<li> See the Clawpack 4.3 documentation [link:claw43user.pdf (pdf)] for
information about plotting in Matlab.

<li> Clawpack 5.0 contains a new approach to plotting using Python, see the
documentation at [claw:doc/plotting] for details on how to use this.

<li> If you want to use the new Python plotting routines to plot results
produced by the old Clawpack 4.3 this should work fine, provided the output
is modified to insert the require ndim parameter in each fort.t file.
This can be done automatically for all such files using

<pre class="clawcode">
  unix> python $CLAW/python/fix_fort43.py
</pre>

in the directory where the old fort.t files are.

</ul>
end_html