When a struct is malloc'd, it is obvious how big that struct is. It is very tedious however to declare 64 dummy dword fields for a struct of 256 bytes. Declaring the struct to have a fixed size array doesn't help much, because you can't name the elements individually. Currently I'm using an AutoHotKey script for adding the fields, which stopped working today for whatever reason.
So: Is there a way to add multiple fields in a struct declaration at once? Maybe via some idc api?
There's a neat, but unintuitive, trick for doing this in IDA which I use all the time: define an array, but in the array dialog untick the "Create as array" tickbox. What this does is create as many elements as you want in the struct, but doesn't make an array out of them.
The workflow is this:
Create a single element of the type you want (e.g. a dword) at the offset where you want your list of elements. (This will often be zero when you're just creating an empty struct that you know the size of, but nothing about the internal structure).
Press the * key with your cursor on the element you just created.
Enter the number of elements you want in the array dialog.
Untick the 'Create as array' tickbox in the lower left of the dialog box.
Hit OK.
IDA will create as many elements as you asked for, but not as an array. You may need to re-type stuff later as you learn what everything is, but this gives you a good skeleton to start from.
I'll give you two solutions. The first may not be exactly what you want, depending on the situation. I can elaborate on either method if you would like.
The cheap method: Using filler arrays
If you have a large structure and there are only a few members whose purpose or size are known, it's often useful to fill in the gaps of the structure by creating temporary dummy arrays. This makes the structure definition more readable and maintainable in IDA, and it also allows you to shape the structure to a specific size without defining more members than you need to.
Let's say you have a structure which is known to be 0x400 bytes in size, and you know the definitions of the members at offsets +0x0 and +0x384. Let's also say you know that there are words at +0x4 and +0x6, but you don't know what they represent yet. Then you might define the structure to be something like this:
00000000 MY_STRUCT struc ; (sizeof=0x400)
00000000 ProcessID dd
00000004 field_4 dw
00000006 field_6 dw
00000008 __filler1 db 892 dup(?)
00000384 ProcessObject dd ?
00000388 __filler2 db 116 dup(?)
000003FC __filler3 dd ?
00000400 MY_STRUCT ends
This is much more readable than it would be if I didn't have the __fillerX elements there. It's also arguably more correct, because you have no way of knowing ahead of time if all the members of the structure are actually qwords.
The reason I have __filler3 at the bottom is because if I ever want to define an element in the region of __filler2, I can wipe __filler2 (and later add new filler members in that space) without shrinking the overall size of the structure. In fact, if the size is the only thing known at the time of structure definition, defining a final element should probably be the first thing you do. Then you'll never need to figure out the sizes of the filler arrays yourself; the default array size that IDA gives you will always be correct. To do that quickly, I usually just create an array of size N-4 and tack on a dword at the end.
But I actually want 32 qwords!
I can't think of any practical situations where this would be useful, but maybe you can!
In any case, you can, in fact, do this through the IDA API. The functions you want are AddStrucEx and AddStrucMember. You should be able to get most of the information you want from the IDA help docs.
If you want a working example of an IDC script to generate structures, you can generate one yourself. Create a few structures, then go to File > Produce File > Dump typeinfo to IDC file....
Or this should also work (uses IDAPython):
id = AddStrucEx(-1, "EXAMPLE_STRUCT", 0)
for i in xrange(0,256,8):
AddStrucMember(id, "field_%x"%i, i, FF_DATA|FF_QWRD, -1, 8)
Related
I've created a Pure Data abstraction which takes 5 MIDI controller IDs as arguments, and has two inlets (velocity and MIDI controller). I use spigot to output the input velocity at outlet i if the MIDI controller ID matches the i th argument to the abstraction.
I use this to control 5 sliders (volume, attack, decay, sustain, release) from my MIDI controller, which look like this:
As you can see, the abstraction is very repetitive, and ideally it would be able to take n arguments, outputting the slider value of the given MIDI slider at the i th outlet. Alternatively I could just use a single argument and use the abstraction anew for each slider, but that's a bit repetitive too.
Is there a way to take n arguments in my abstraction? Or, is there any other more elegant solution to control sliders from MIDI that I'm missing?
For core vanilla objects or externals it is a common feature to take a variable amount of arguments. Many objects like [trigger], [select], [pack] or [unpack] behave this way and create a number of xlets according to how many arguments they are given.
For abstractions however, there is no simple way to have a variable number of outputs depending on the number of arguments given. It's possible, but it's rather a hack.
Let's first look at some techniques to make the patches easier.
As you can see in the following screenshot the [expr] object in your patch can easily be substituted by a simple Pd object: [==]
Let's look at the MIDI messages. MIDI objects output the elements of the hierarchical tree: channel, controller, value in this order but from right to left (true for all objects in Pd, the rightmost outlet fires first). What you want is routing them according to the controller number, ignoring the channel number. For this you need to flip the order of the two first outlets so you can use the controller number as a selector in a list. One way to do this is to pack the numbers and then reorder them in a message using the dollar symbols (careful, dollar symbols in message objects operate on local messages, they have nothing to do with what dollar symbols do in object boxes). Another method would be to use the object [swap] (using [swap] in this case would reduce the number of objects by one in the patch, but the [pack] and [message( is more versatile, that's why I show it here).
Regarding the actual question of taking n arguments to change the number of outlets: This challenge are two separate problems:
How to detect the number of arguments
How to change the number of outlets dynamically
Until recently there hasn't been a way proper way to detect the existence of arguments. Since Pd 0.50 there is [pdcontrol] which allows exactly this.
Using [initbang] here because [loadbang] would fire too late for creating the outlets.
The second part of the challenge may be solved either through a technique called 'dynamic patching' or by outputting a list instead of creating additional outlets. However with a list output you'd need to route it afterwards and then you'll arrive quickly at the patch pictured above.
I have a process which is repeated on a set of data stored in separate folders. Each time a certain folders data is processed I need new variable names as I need to results separate after the initial processing is finished for more processing.
For example at the start of each new block of the repeated function I declare sets of arrays
Set_1 = zeros(dim, number);
vectors_1 = zeros(dim, number);
For the next set of data I need:
`Set_2 = .........`
and so on. There is going to be alot of these sets so I need a way to automate the creation of these variables, and the use the new variables names in the function whilst maintaining that they are separate once all the functions are completed.
I first tried using strcat('Set_1',int2str(number)) = zeros(dim, number) but this does not work, I believe because it means I would be trying to set an array as a string. I'm sure there must be a way to create one function and have the variables dynamically created but it seems to be beyond me, so it's probably quite obvious, so if anyone can tell me a way that would be great.
I'd not do it like this. It's a bad habit, it's better to use a cell array or a struct to keep multiple sets. There is a small overhead (size-wise) per field, but it'll be a lot easier to maintain later on.
If you really, really want to do that use eval on the string you composed.
The MATLAB function genvarname does what you want. In your case it would look something like:
eval(genvarname('Set_', who)) = zeros(dim, number);
However, I would follow the recommendations of previous answers and use a cell or struct to store the results.
This sort of pattern is considered harmful since it requires the eval function. See one of the following for techniques for avoiding it:
http://www.mathworks.co.uk/support/tech-notes/1100/1103.html
http://blogs.mathworks.com/loren/2005/12/28/evading-eval/
If you insist on using eval, then use something like:
eval(sprintf('Set_1%d = zeros(dim, number);', number))
I have multiple vectors (+100) that have been loaded into MATLAB workspace, I would like to write a script that can plot and save them all, but for that I need their name, my question is: is there a way to automatically get the name vectors saved in the workspace.
thanks in advance.
Step one: whoever gave you a *.mat file with 100+ named variables in it, [censored for strong language and scenes some viewers may find upsetting]. I am only partly joking here; if you find yourself in this sort of situation normally it is because something has gone terribly wrong upstream. We can work around it, though.
Step two: use who with the filename to get a list of variables in that file
names = who('-file', 'all');
Step three: load the variables (or a subset of them) into a struct
data = load('all.mat');
Step four: use dynamic structure naming to extract data:
for n = 1:length(names);
plot(data.(names{n})); % or whatever you want to do with this data
end
I would probably just use the loop to dump the data in a cell array so as to make further processing simpler and avoid further use of dynamic field names or worse, eval.
You can use who, which lists all variables alphabetically in the active workspace.
I'm reasonably new to Matlab, and have been trying to teach myself. I have looked for a similar question, but can't find one that's quite right.
In my workspace I have several structures with similar names. These structures will always start with the same word ('Base'), though the rest of the name will change ('1', '2', '3'), so for example Base1, Base2, Base3... etc. These variables were generated using the data cursor tool in a figure, so contain the fields Target, Position and DataIndex. I am only interested in the value in Base*.Position(1,1). I would like to extract this value from each structure, as many times as there are structures (in one instance there may be 6 structures, another time only 4).
I am considering using the eval function, but it seems to work on exact strings rather than only the first part of a name. Additionally, a lot of documentation seems to advise against using eval.
So far I have:
clearvar except 'Base*'
list_variables=who;
for i=1:length(list_variables)
BaseTS(i) = eval('Base1.Position(1,1)');
end
It's the for loop I'm stuck on, as I don't know how to generalise so it will extract the value .Position(1,1) for each different structure name.
Thanks in advance
Instead of having many structures called Base1, Base2 etc rather put your structure in an array. Then you could rather call Base(1).Position(1,1), Base(2).Position... etc. Your code will be more flexible and manageable this way.
So I suggest when you export using the data cursor, export to a variable called Base_temp and then immediately stick this into the next element of an array:
Base(end+1) = Base_temp
or even:
Position(end+1) = Base_temp.Position(1,1);
Then it's just a case of pressing up and enter after each time you export with the data cursor.
What you have read about avioding eval is correct, it's very rare (if ever) that eval is a good idea. It makes your code hard to read and very hard to debug. But since you're learning, this is how you could fix your loop. (But don't do this way, seriously don't, use arrays rather):
for i=1:length(list_variables)
BaseTS(i) = eval(['Base', num2str(i), '.Position(1,1)']);
end
in other words use string concatenation to build up your string and use the looping variable (i) to get the different numbers. You'll need num2str to convert fromthe number to the string. But don't do it this way. This is a bad way.
Dan's suggestion about avoiding eval is very valid. But if you decide to keep on the structures you have in your workspace, here's something without loops, but again cellfun seems to use loops internally. So, I guess this could be an alternative solution, with the not-so-popular eval -
list1 = who('Base*')
list2 = cellstr(strcat('BaseTS(',num2str([1:numel(list1)]'),')='));%%//'
ev1 = strcat(list2,list1,'.Position(1,1)');%%//'
cellfun(#evalc,ev1,'uni',0)
I am looking for a way to handle different type of data in one column of a Matlab uitable.
Usually uitable is used to set whole columns to the same data type, such as logical (gives a checkboxes), char (gives left-aligned text), numeric (gives right-aligned number) or a 1xn-cell-array (gives pop-up menus with different choices).
It is set using the columnformat property of uitable, e.g.
columnformat = {'numeric', 'logical', {'Fixed' 'Adjustable'}}; % 3 columns
You can find an example at matlab documentation.
I am looking for a way to set the format for single cells to realize something like this:
Matlab's uitable is a crippled version of an underlying JIDE table.
It is possible to get to the underlying java (see findjobj in file exchange), but that requires a lot of work. Yair Altman's undocumented matlab site is a good starting place for understanding the java side of matlab.
It sounds like you want something like a property editor, as opposed to a generic UI table -- i.e. properties listed in first column, property value editable in second column. There are a few "off the shelf" versions in the file exchange, which use JIDE:
See propertiesgui, or property-grid for mostly functional examples. The second example is easier to use -- you simply provide a class or struct, and it creates the proper field entry format. The first one offers more choices -- like color boxes, drop downs, etc, but requires you to be more involved in specifying how things behave.
I had the same problem, but in the end it worked by giving the (numeric) cell an (char) initial value. When changing the char value from the UI the format of the cell remained char, although the rest of the column was numeric.