| Re: Need ideas for a data structure [message #3073] |
Fri, 04 November 1994 17:22 |
mark_cadwell
Messages: 10
Registered: February 1994
|
Junior Member |
|
|
I don't know about IDL, but in PV-Wave I created a flexable record type
that works similarly to what you describe.
There is a set of non-changing header elements with things like titles,
axes annotation, data start and stop times, etc. There is also an element
that contains the size of the data arrays that follow. The file is opened
and read into PV-Wave with a READU function call. You have to create a
data structure in which to put the stuff you read. You can define one
data structure that has everything up to and including the element that
tells you array sizes.
Once you have that information, use it to create a second data structure
that has all the header stuff AND arrays of the proper size. Rewind the
data file with the function "POINT_LUN, UNIT, 0". This sets the I/O
pointer back to the beginning of your file so you can read it again (this
time with your data arrays).
It may not be the most elegant solution, but it has worked very well for
me. The only thing you have to look out for is that the data arrays must
start on a fullword boundry. If you need to, you can pad your data
structures with a few empty bytes to get to the fullword boundry.
--
----------------------------------
mark_cadwell@qmail4.trw.sp.com
|
|
|
|
| Re: Need ideas for a data structure [message #3087 is a reply to message #3073] |
Thu, 03 November 1994 03:04  |
steinhh
Messages: 260
Registered: June 1994
|
Senior Member |
|
|
In article <3992q9$2pp@aplcomm.jhuapl.edu>, CroucAR1@subtech1.spacenet.jhuapl.edu writes:
|>
|> I'm trying to devise a way of dealing with a bunch of data items which will
|> consist of a fixed number of header fields and a widely varying number of data
|> entries. For one such item, an anonymous structure seems appropriate, but I
|> don't see how I can generalize this to an array.
|>
|> If I define an array of structures based on the largest array (which I won't
|> know ahead of time anyway), I'll waste a huge amount of memory.
|>
|> I thought of building an array of pointers to anonymous structures, but the
|> structures seem too anonymous for this to work.
|>
[..]
|> PS: I'm bi - it needs to work with both IDL and PV-Wave
I was just about to propose IDL handles, but then I saw your PS, which
makes it slightly more difficult.
But despair not! The solution is to do your own memory management --
I know it's not pretty, nor very efficient, but it works!
I gather that you have a data structure that could be represented
by the following "conglomerate":
Fixed info
-----------
| | ____ --------_____-------- ___ ...... (N entries)
| | |Entry | |Entry |
| | | | | |
----------- -------- --------
and you'd like to store an array of such data conglomerate.
The fixed info is easily put into an array, with two tags, "start"
and "length", in addition to those representing fixed data.
Assuming that each "Entry" is of the same data type -- this is crucial --
create a global (common block) array of, say, a thousand entries, and
associate with it a global counter Nused (initialized to zero), keeping
track of how much of the array that's in use.
Now, each time you need a new conglomerate data sturcture, create the
Fixed info part. You ought to know by this time how many entries are
needed, so initialize the "start" tag in the fixed info with the current
value of "Nused", and increase Nused with the number of entriest that are
reserved for the new data conglomerate.
If the entries are also variable, you need to go one step further down
into the data structure, and implement the same idea there...
Stein Vidar
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
