variable coding help

[aaaaamartin]

Hello,

I'm looking for a way to assign a value to a variable as shown below.
Note that I used long for "var" ...
What I do not want is: 1) waste ram, 2) waste rom, 3) no solution ;-)
I have PCH 3.235.

[PCM programmer]

See this thread for a discussion of switch-case code efficiency:
http://www.ccsinfo.com/forum/viewtopic.php?t=21373&highlight=switch+case

[aaaaamartin]

Thanks PCM,

I see, switch-code is efficient.

But, let me go a bit further into detail...

That's what I'm doing:

The Pic can recieve dumps of data, from a host, that it needs to assign to certain variables.
The host of course, does not worry about any internal variable names and it references to these with an address.
A packet recieved therefore looks like: [address - value] * n.
The Sender does not take care about the order of these pares among each other,
In other means, it does not necessarily need to update all of these variables.

Further, the host can requeset the pic to send back the value of a variable given an address.

Finally the host can request the pic to dump all of it's variables together with the address as the host knows them.

I need some hints, to not end up in a mess.

Thanks Martin

[Ttelmah]

Questions:
How many variables are involved?.
What are the sizes of the variables (all the same, as shown, or different for different commands)?.
Are the commands 'sequential' (in that you get a 'number' - var in your example that goes through a limited range, without gaps)?.
How many possible 'var' values exist?.
The normal way for a small number of values as shown, would be to simply use a array. If you needed local 'names', rather than using numeric values, then use a union.

Best Wishes

[aaaaamartin]

[Ttelmah]

The key about the union, is that it allows handling of the different types.
If you declare a 'structure' that contains the data you require, with the names you require, and then a union, of this structure to a byte array, you can access the data either 'by name' through the structure, or as bytes, in order through the array.
Now, I do a thing like this, with a system I have. I have five different 'packet' forms, sent back to the host computer, each with different layouts. I have five structures defining these layouts, used internaly by the PIC code, and a union between these and a byte array. My parser, returns a number (keep this much smaller - if you are only dealing with 100 values, limit it to perhaps 8bits), which is then used as an index to an array, which contains the address in the byte array of th data required, and a second array, which contains the sizes. So in my case (for example), value 32, comprises the address in the union, of an array of four floating point TOC values, and the corresponding 'size', is 16. The transfer program simply transfers the bytes (adds a checksum, etc.), and doesn't care what the data actually is. It simply knows it has to send 16bytes from the 'data' block for this command.
In memory, the same area, is accessed using the structures, allowing the PIC code to talk to it by name etc..
Making your address range larger, just increases code size. If you want it larger to allow multiple units to be used on the same bus, then have the large address comprise a smaller 'local' address, with the upper bits giving the unit address.

Best Wishes

[aaaaamartin]

Ttelmah,

I must admitt, I don't get it to work.

Can you please show an example.

Thanks Martin

[aaaaamartin]

Hello,

I now use a const array of pointers to each var and another arrray,
telling about the size of var, because I use int pointers.

Anyway I would like to see an example of Ttelmah's approach.

Thanks Martin