Re: InFile method usage

[Marcus G. Daniels]

>>>>> "RLR" == Rick Riolo <address@hidden> writes:

RLR> 1. Does a method like getDouble (&friends) span white space
RLR> (blanks,tabs,newlines) before deciding if the next token is a
RLR> double, and if so, loading it into the given double variable?

Yes.

RLR> 2. Does getDouble accept values like 22 32 4444 as valid double
RLR> values?

Yes.

RLR> 3. If input looks like: 333.3 abc 4443.3 and I send three
RLR> getDouble messages to anInFile, will it read 333.3 ok, return
RLR> error for abc, then read 4443.3 ok?  That is, does it leave the
RLR> InFile read-point after the "abc" after an error?

It leaves the read-point before the abc.  This is the behavior of scanf
conversion failures.  Thus, the 4443.3 will never read.


To load data in to Swarm, I suggest using R
(http://www.ci.tuwien.ac.at/R) to read in the data, and then store it
as HDF5.  R is programmable, so there are many options for preprocessing
data for Swarm. 

Here's an example: 

1. Given a text file of data to load, say `input.txt',
----
$ cat > input.txt << EOF
val1 val2 val3
222  32   4444
111  16   2222
EOF
----

2. First, run R, the version at:
 ftp://ftp.santafe.edu/pub/swarm/support-software
----
$ R
----

3. Read in the data file from input.txt, using the first line to define
the field names.  Assign the result to the `data' variable.
----
> mydata <- read.table("input.txt", header=TRUE) 
----

4. Then, tag the `data' variable with attributes telling Swarm how to 
use it.  Tell Swarm that the object should be loaded as a List.
---- 
> attr(data,"type") <- "List"
----

5. Also tag the object telling Swarm how to treat each component in
the List.  Tell Swarm that the items in the List should be of
MyClass type (see the class definition below).
----
> attr(data,"component-type") <- "MyClass"
----

6. Currently, in the simplest case, Swarm expects to get
objects from two sources.  One place is the ~/.swarmArchiver file,
e.g. the place where window geometry is saved.  The other place, new
to Swarm 1.4, is from a file called ~/swarmArchiver.hdf.  Objects
stored in both of these files are organized hierarchally by the owning
application, and the application's `personality' or `mode' identifier.
In order for a Swarm model to read an object from one these files, it
is necessary to tell Swarm what model will own the object.
----
> inputapp <- list(default=list(data=mydata))
----
What this expression means is to create an object of the form below, and
store it to the variable inputapp.  (The other names, default and data,
are not global names, but exist only within their respective lists.)

+------------------+
| inputapp         |
|  +------------+  |
|  | default    |  |
|  | +-------+  |  |
|  | | data  |  |  |
|  | +-------+  |  |
|  +------------+  |
+------------------+

The `inputapp' and `default' shells around `data' is necessary for Swarm
find what data is relevant to a particular application mode-of-use.

7. Finally, once you've created the data, and described the application
context, you can save it to file: 
----
> hdf5save("swarmArchiver.hdf", inputapp)
----
(This is assuming you are in your home directory.)

8. Quit R.
----
> quit()
----

9. Now the data can be loaded in to Swarm.  Although Swarm will create
classes on-the-fly if necessary, it tends to be easier to work with an
object if there are some methods defined, although Probes are another option.
The data file above is described by a class like this:

----
@interface MyClass: CreateDrop
{
  double val1;
  double val2;
  double val3;
}
@end

@implementation MyClass
- print
{
  printf ("%f %f %f\n", val1, val2, val3);
  return self;
}
@end   
----

10. To recover the List of MyClass instances stored in the HDF5, you
simply ask for the object by name:

----
 obj = [hdf5Archiver getObject: "data";
----

11. Then printing the results:

---
 [obj forEach: M(print)]; 
---

..you get:

----
address@hidden $ ./inputapp
222.000000 32.000000 4444.000000
111.000000 16.000000 2222.000000   
----

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.