fdep: a pedestrian function dependencies finder

FDEP dissects ML files and iteratively looks for all user defined functions used during runtime

Description

FDEP dissects MATLAB files and iteratively looks for all user defined functions (modules), which are used during runtime

FDEP retrieves for each module its
- subfunctions
- nested functions
- anonymous functions
- eval class calls
- unresolved calls
    and all
- ML stock functions
- ML built-in functions
- ML classes
- ML toolboxes
    that it uses

runtime options and returned macros create user-friendly,
intuitively comprehensible, and interactive GUIs, which
- list the results in various panels and listboxes
- plot a full synopsis for each module with
exhaustive information
- plot a full dependency matrix
- plot a runtime and modules tree

in essence, FDEP is a wrapper for DEPFUN and MLINT; however, due to an efficient pruning engine it is considerably (20-100 times!) faster

FDEP is particularly useful if you want to distribute your functions and need so see what else to include as well as the toolboxes they use

FDEP may be useful to debug an erroneous function with correct syntax

see
> help fdep
and the accompanying published M-file for help and an exhaustive example

Acknowledgements

Farg: A Pedestrian M File Parser Showing All Used Functions' Syntax and Detab: A Pedestrian String Detabulator inspired this file.

This file inspired Ftoc (V1.2), Graph Viz Like Tools For Matlab, and Dep.

• MATLAB release: MATLAB 7 (R14)
• Other requirements: optimized for R2010+

[Matt Wolinsky]

Very useful tool, and fast!

It worked great for me, except for one issue: if the input function calls subfun.m only from within an anonymous function, then fdep() does not find it.

In other words if myfun.m has the lines

Sfunc=@(x)subfun(x);
x=rand;
y=Sfunc(x);

then for p=fdep('myfun'), p.fun does NOT contain subfun.m

This came up because I tend to use anonymous functions as aliases a lot, e.g. Fab=@(c,d)F(a,b,c,d)

[Paul]

WOW! brilliant and fast. Highly recommend! Thank you

[Reza Farrahi Moghaddam]

 

[Justin Elstrott]

 

[Justin Elstrott]

 

[Jaroslaw Tuszynski]

exactly what I needed

[Adam A]

 

[Alvaro Delgado]

 

[Nilimb]

very helpful and fast.

[Eric]

Great function. Saves lots of time and headache. Much faster than depfun which I had to force quit to stop after 20 minutes.

[Kelly Kearney]

This is an excellent function that I've used for years. However, the GUIs don't play nicely with multiple monitors. Because of the mix of normalized and absolute units used for various parts of the GUI, simply resizing the whole thing post-render isn't ideal (and I just discovered by accident that there are some buttons in the upper corner that I never knew existed in all my time using it). It would be nice if I could somehow tell it which portion of the screen to use (even if I had to hardcode that in my own version, but right now that seems to require changing several lines throughout the code, and I haven't tracked them all down).

[Aurelien Queffurust]

WoW ! which a great tool!

[Ben Petschel]

Excellent tool, much faster and more informative than DEPFUN.

An especially useful feature is detection of unresolved function calls, which DEPFUN doesn't seem to do. The synopses of each module have a list of unresolved names, though I couldn't see a way to get a single list of all unresolved names except by browsing.

Also you might want to switch off latex interpretation in the point labels in the dependency matrix (line 1921, I think).

Good work!

[Jon Griffiths]

Hi, I am having some problems using fdep on a script/function that calls functions which are part of a collection.

e.g. fdep('myfunction', '-m')
where
function [x] = myFunction(y)
x = mycollection.getX(y);
end

Is there a switch to capture this?

Thanks
Jon

[Kassi]

This is awesome! My job requires using/debugging code written by someone else and this is extremely useful. Thank you for easy to navigate, clean, well written code :)

[Shahar]

Amazing!

[AMVR]

(Continuation of my previous error report). In case it helps, after tracing the error I reported earlier, I noticed the following:

The error occurs at:
ixb(1:par.mfun,1)=find(ixt(:,1)==1);
(??? Subscripted assignment dimension mismatch.)

and this is because

K>> par.mfun
ans =
32

K>> numel(find(ixt(:,1) == 1))
ans =
41

[AMVR]

I'm getting the following error when running fdep on a function that uses IBM's CPLEX
26 7 : | cplexbilp P 1: C:\ILOG\CPLEX_Studio_AcademicResearch122\cplex\matlab\x64_win64\cplexcreateprob.p
27 8 : | cplexcreateprob R P 1: C:\ILOG\CPLEX_Studio_AcademicResearch122\cplex\matlab\x64_win64\cplexoptimset.p
28 9 : | cplexoptimset P 1: C:\ILOG\CPLEX_Studio_AcademicResearch122\cplex\matlab\x64_win64\@Cplex\Cplex.p
??? Subscripted assignment dimension mismatch.

Error in ==> fdep>FARG_get_entries at 3066
ixb(1:par.mfun,1)=find(ixt(:,1)==1);

Error in ==> fdep>farg at 2585
[p,par]=FARG_get_entries(p,par);

Error in ==> fdep>FDEP_get_fun at 770
[fa,fap]=farg(fnam,'-s','-d');

Error in ==> fdep>FDEP_get_dep at 664
[p,par,dtmp,dmlf,dmod,dmcl,docl]=FDEP_get_fun(p,par,fnam,frot);

Error in ==> fdep>FDEP_get_dep at 740
[p,par]=FDEP_get_dep(p,par,dtmp{i});

Any ideas why this happens?

[CLF]

Very Useful. Would it be possible to handle also comments of, e.g.
%ToDo-things?

[Julian]

Splendid! I am really impressed with this tool, especially, as others have commented, the speed up over depfun. Do you know if it will work with the new(ish) user classdefs?

[Ole]

Ok, I have been damn stupid. Sorry. This piece of software is excellent!

[Ole]

Thanks a lot, generally it works great! Unfortunately I have to analyze a gui-based software project and fdep seems only to consider automatically called functions. E.g. submodules which are called due to user input are not analyzed. Is that correct?

And if yes, what can I do about it? I really need to find out, which toolboxes are involved in a rather confusing project I recently "inherited".

[Andrew]

 

[us]

andrew

unfortunately: no...
one of the reasons why the .EVAL. group is collected separately by fdep and a warning flag is displayed if any such construct is found...

which, in turn, is one of the reasons why one should follow these rules
- do NOT use EVAL
- use FEVAL with function handles, only
note:
feval(@sind,30)
will correctly find the call to SIND...

sorry
us

[Andrew]

Is fdep capable of finding feval calls, where the function name argument in feval is itself a string variable?

e.g.

feval(fname,1,2,3)

[us]

thomas
thanks for your comment

to achieve what you need, you could do this

pp=fdep('yourmain');
% now,
pp.sub.M
% is an array listing all the modules with their full path
% you could simply create a little utility, which does the copy, eg
function copymod(pp,newfolder)
% error checking etc left out...
for i=1:pp.nfun
copyfile(pp.sub(i).M,newfolder);
end
end

us

[Thomas Benson]

This is fantastic. It worked quickly on a matlab program that contains 184 functions. An additional capability that would be useful would be to have an option in fdep to copy all the non-toolbox dependencies to the current working directory, although I think I could write this myself from the information provided in the P structure. Thanks us.

[Peder Schmedling]

 

[Oleg Komarov]

I really find useful this tool.

[us]

oleg
thanks for your comment

there are two reasons
1) a MODULE is a M-file/-script, and the first (main) function
is simply another (sub-) function within this file
2) the first function should(!) have the same
name as the M-file (otherwise, mlint complains); this is a simple
sanity check for the user

also, note the first char in the panel
M (...) ... % <- the main function
S (...) ... % <- the first (official) subfunction
S (...) ... % <- and so on

us

[Oleg Komarov]

The package is great but I don't understand why the main function is listed in the subfunction pane (or in the 'sub.S.fn'):

function foo
end
function foo2
end

I'd expect only foo2 to be there. What is the rationale?
Tnx

[Jonas]

Excellent function!

In Windows, the file separators in output.froot point the wrong way, though. Is that on purpose?

[us]

hi dirk
great suggestion for an enhancement!...
i built a helper function, which does what you need and will email it to you if you want...
it will become part of FDEP as soon as i can get to it...
us

[Dirk]

Thanks for this fantastic tool. The speed-up compared to depfun is amazing.

Though there is one feature I am missing somewhat:
fdep lets you identify which toolboxes are used from _somewhere_ in you code easiliy. But unfortunately it is not so easy to identify which of my .m files caused this toolbox dependency.
What would be great would be a possibilty to, e.g. select one of the toolboxes in the list-view which in turn could highlights all modules that directly depend on the selected tool box.

[us]

jason

yes, this is by design...
while it could easily be changed, it also (potentially) would lead to undesirable results...
thus, to FDEP a file residing in a folder, which is not in ML's path, you must CD into that folder...
however, let me know if this is a real problem for you, as it certainly can be changed...

us

[Jason Kaeding]

For being as full-featured as it appears, why will this function not accept full path names?

fdep('C:\blah\blah\blah\file.m')

I get an error saying "file not found". It only seems to work with files in current directory or in the MATLAB path.

[Andreas]

I liked the previous version very much! Thanks a lot.

The latest version has a bug. If a sealed class (e.g. with a definition "classdef (Sealed) MyClass") is analyzed, the program crashes in fdep on line 2928. If "(Sealed)" is removed the program works as advertised.

[us]

with the latest version (08-Jun-2009 22:24:13/R14-R2009a), per isakson's error messages do not show anymore...
the reason was an old parser engine, which did not handle more recent function constructs...
thanks, per, for alerting me to this problem...
us

[Alon]

This is a great idea, implemented in a great way!
Full of user-fraindly and usefull features.

Great job.

[Jveer]

very nice. thank you for sharing this very useful too

[Yuri Kotliarov]

Great function! Very useful. Is it possible to see also other files required in a package, like mat, txt, ini, etc?

[Image Analyst]

I just downloaded it and it looks very nice. However I'd like you to fix the bug where it's locating the wrong module. I have a main module and it calls 8 other modules, 7 of which are in the same folder as the main module. However, module #2 you list as being one of the same name that I happen to have in another folder. I'm not sure why you're seeing that one first - MATLAB doesn't, it sees and runs the proper one which is in the same module as the main module. For example, in folder1, I have m-files myFile and mySubFile1, mySubFile2, etc. But in folder2, there is also a mySubFile1 and this is the one that your program is finding instead of the mySubFile1 in folder1.

A few more suggestions on the list GUI:
1. Replace the [John D'] button with a scroll bar or two small buttons to increase and decrease the font size.
2. I'd also like to see tooltips on your 6 buttons (right off the bat, it's not obvious what some of them do).
3. I'd like it to start with a GUI rather than a 2-steps command line process.
4. The Image Processing Toolbox is incorrectly called "images" in the toolboxes listbox.
5. A longer and more descriptive name than the short and cryptic "fdep" would be nice. And what does "pedestrian" have to do with it? Is that word needed?

Here's what MATLAB reports for my toolboxes:
MATLAB Version 7.7 (R2008b)
Image Processing Toolbox Version 6.2 (R2008b)
MATLAB Builder NE Version 3.0 (R2008b)
MATLAB Compiler Version 4.9 (R2008b)
Signal Processing Toolbox Version 6.10 (R2008b)
ToolBox EXAMPLE_SCRIPTS Version 1.0.0

Just my 2 cents on a well done program.

[exfnjbdr exfnjbdr]

[URL=http://fxbpibjx.com]guupplld[/URL] <a href="http://hjnyyjjr.com">jmtcnixm</a> zshmaoig http://spqkqjpu.com jqpjujjw apsdptbx

[Urs (us) Schwarz]

andrew, can you get in touch with me?unfortunately, you didn't provide an email address.

[Andrew Walton]

It falls over when mex files are used.

[urs (us) schwarz]

to accommodate FDEP to JD's eyesight, an upgrade was just uploaded with these changes

1) candy colors were moved towards zero (still hoping that they group the various panels quickly for the eye...)

2) a JOHN D' button was added; its help section says:
utility in honor of John D'errico, a senior and most respected
FEX and CSSM contributor, with very poor eyesight
CLICKING on the button will grow the fontsize by 1 point every time

the version tag is 07-Nov-2007 21:10:00
us

[John D'Errico]

As always, a fully featured, fully able tool from us. Well done.

I do have one comment - my older eyes had a moderately hard time with the tiny font size and color scheme in the resulting gui. Pastel colored text on a pastel background was hard to read. Time for a new prescription in my glasses?

[Caroline Herbst]

Extremely helpful, indeed! Also, the author did very quickly add a new feature after a pleasant email exchange.

[The empror of Matlab]

wow!