I'm working on a static library project for a c++ course I'm taking. The teacher insists that we define only one function per source file, grouping files/functions belonging to the same class in subdirectories for each class. This results in a structure like:
MyClass
\MyClass.cc (constructor)
\functionForMyClass.cc
\anotherFunctionForMyClass.cc
OtherClass
\OtherClass.cc (constructor)
Whether this is good practice or not is something I'd not like to discuss, since I'm simply obliged to organize my project in this manner.
I'm working in visual studio 2008, and somehow got strange link errors when using an identically named function (and thus filename) in two classes. This appears to be caused by the fact that visual studio puts all .obj files (one for each source file) in one intermediate directory, overwriting earlier generated object files when compiling identically named source files.
This could be solved by putting the object files in subdirectories based on the relative path of the input file. Visual studio allows one to configure the names of object files it generates and has macros to use in there, but there appears to be no macro for 'relative path of input file'.
So, is there some way to get this to work? If not, is using one project for each class the best work-around?
You are right, by default all object files are put into the same directory and their filenames are based on the source file name. The only solution I can think of is to change conflicting file's output file path in here:
Project Properties-C/C++-Output Files-Object File Name http://img37.imageshack.us/img37/3695/outputfile.png
PS. It sounds like the lecturer has a crappy (probably written by the lecturer himself) automatic code verifier that imposes this restriction. To get extra marks, offer to rewrite the parser so it works with normal/sane/non-weird projet layout.
Real answer:
Change
C/C++ => Output Files => Output File Name
to
$(IntDir)/%(RelativeDir)/
Every .obj file is going to be created in a sub folder so its not going to overwrite the previous on linking.
I can't think of any way to fudge the project settings to get VStudio to automatically split out the intermediate files into separate folders.
You have a few chances -
Build the class name into each file name. Most IDE's display just the file name in the tab view so if you do have several methods in different classes with the same name, its going to be difficult to tell them apart if the file name does not include the class name along with the method name. Which is really why I think your teachers advice is madness. I have not seen any programming style guide advocating that approach. Additionally it goes directly against the way various tools work - if you use Visual Studio to create a class, it creates one cpp file and one header, and automatically appends each new function to the single cpp file.
You could create a static library per class. When linking in static libs the obj files are all packaged up inside the .lib so conflicts are no longer a problem.
Switch comp-sci courses to one thats not being taught by a nut job. Seriously, this guy is completely out of touch with industry best practices and is trying to impose their own weird ideas on their students: Ideas that are going to have to be unlearnt the moment they leave the teaching environment.
You can also change output file name per file in its properties. Just make sure you use different names.
Can you use the class name in the filename to disambiguate? I'm thinking that you might have
MyClass
\MyClass.cc (constructor)
\function1_MyClass.cc
\function2_MyClass.cc
That would mean that every file would have a unique-enough name to defeat the problem. Is that an acceptable strategy?
You could probably arrange the properties of the project to put the object files into a folder which is below the folder of each source file. Once the project has this property, then every source file should inherit this property. (But if you've done experiments like Igor has suggested, then you may need to go through the properties as reset them back to the parent).
Having looked at the help files, I think you should go to project properties/C C++/Outpuf Files/Object File Name: and enter $(InputDir) (no trailing backslash). Every source file should then inherit this property and your .obj files should be separated.
You may need to do a Clean Solution before you make any changes.
Renaming the object files will work, but it's going to be a pain, and it will slow your compile/link cycle down. I've never figured out why, but it seems to confuse Visual Studio if the object files don't have the default names.
You could prefix the funciton name with the class name; e.g. myclass-ctor.cc, myclass-function1.cc etc.
You could have one .cc file per class which #includes the individual function files. In this case you'll need to prevent the #included files from being compiled seperately (either rename their extension or set Properties->Exclude From Build to 'Yes').
Out of curiosity, where does your teacher want you to put free functions e.g. local helper functions that might normally belong in an anonymous namespace?
If not, is using one project for each class the best work-around?
Not a good idea - apart from the fact that you won't end up with a single static library (without even more jiggery pokery), your link times are likely to increase and it will hide a
lot of pertinent info from the optimizer.
On another note; If the course is actually about C++ not OO programming, do what you need to pass but take your teacher's advice with a pinch of salt.
You don't have to put them in different translation units... why not put each function in a .h and include them all in one .cc per class? That will very likely give better output from the compiler.
I'd be asking why the teacher is insisting on this odd structure, too, the reasoning behind it should be explained. I know you didn't ask that of us, so that's all I'll say.
In Visual Studio 2010, I set
Properties -> C/C++ -> Output Files -> Output File Name
to
V:\%(Directory)$(PlatformName)_$(ConfigurationName)_%(Filename).obj
for OBJ files to end up next to the sources assuming the project lies on drive V (no idea whether there is a macro for it, yet).
By the way: $(InputDir) refers to the solution/project directory and will cause the same problem in another directory.
Related
I'm taking over a project from a coworker that involves several extensive SAS process flows. I have all the files with all the same names and a copy of the process flows they used. Since the file paths in their processes are direct references to their computer, normally I would just re-import the files with the same output names and run the process from there. In a few cases I would have to recreate a query builder as I'm using a few .sas7bdat files from another project.
However, there are quite a few files involved and I may end up having to pass this to another coworker in a few months, and since I can't get a good look at exactly what the import task is doing I'm concerned I may have some of the variables imported incorrectly. Is there an easy way to just change the file path the import or other task refers to?
Given the updates in comments, there's two possibilities I see.
If the paths you're changing are, or can be, relative to the location of the EGP, then you can right click on the Project->Properties->File References and check "Use paths relative to the project...", which means instead of storing a file in c:\my EGP folder\my code folder\code.sas it would store it as my code folder\code.sas. So then if the whole project moves to another computer (or just any other folder) then it automatically has the right path. This is mostly useful for code or similar things.
Otherwise, you're going to have to convert things to SAS code modules. There you can use macro variables to define the locations of things.
I've had trouble with this issue across many languages, most recently with C++.
The Issue Exemplified
Let's say we're working with C++ and have the following file structure for a project:
("Project" main folder with three [modules, data, etc] subfolders)
Now say:
Our maincode.cpp is in the Project folder
moduleA.cpp is in modules folder
data.txt is in data folder
moduleA.cpp wants to read data.txt
So the way I'd currently do it would be to assume maincode.cpp gets compiled & executed inside the Project folder, and so hardcode the path data/data.txt in moduleA.cpp to do the reading (say I used fstream fs("data/data.txt") to do so).
But what if the code was, for some reason, executed inside etc folder?
Is there a way around this?
The Questions
Is this a valid question? Or am I missing something with the wd (working directory) concept fundamentals?
Are there any methods for working around absolute paths so as to solve this issue in C++?
Are there any universal methods for doing the same with any language?
If there are no reasonable methods, how would you approach this issue?
Please leave a comment if I missed any important details with the problem's illustration!
At some point the program has to make an assumption where the file(s) are. Either by getting it from user input or a relative path with the presumed filename. As already said in the comments, C++ recently got std::filesystem added in C++17 which can help you making cross-platform code that interacts with the hosts' filesystem.
That being said, every program, big or small, has to make certain assumptions at some point, deleting or moving certain files is problematic for any program in case the program requires them to be at a certain location under a certain name. This is not solvable other than presenting the user with an error message etc.
As #Hatted Rooster said, it's not generally solvable for some arbitrary file without making some assumptions, however there are frameworks that allow you to "store" some files in the resources embedded into the executable (or otherwise). Those frameworks would usually allow your to handle such files in a opaque way, without the need to rely on a current working dir or relative paths.
For example, see the Qt Resource System.
Your program can deduce the path from argv[0] in the main call, if you know that it is always relative to your executable or you use an absolute path like "C:\myProgram\data\data.txt".
The second approach works in every language.
I have been writing code for a little while now and I recently found out how to create classes in different files and include them in main, along with more cpp files including the definitions of those classes. I was wondering when this is really needed, my code isnt normally that long. Should I use this now as a beginning when my code is only a few hundred lines or less or are the multiple files used with alot more code. In cases like this with such a short code I could probably find it easier just to stick to the main cpp. Any thoughts on this?
Bad habits are easy to learn and hard to unlearn. Why not do it the right way from the beginning?
I advise for the sake of organization that you should create a new set of files - .h and .cpp - for each new functionality you want to add. That doesn't mean create a new file for ever class you want to have. As an example, you can define a Sound class and a SoundManager class in the same header if you want. It's easier to find and edit a faulty piece of code when it isn't all jumbled up with other code in a catch-all file.
When you get to multi-thousand or even multi-million line projects (which I'm sure you will sooner or later), you will obviously be in for a world of pain if you shove it all into one file. Like David Schwartz said, it's better learn it correctly the first time than to relearn it correctly the second time.
If you have different parts of your code in different files, then you can open each file in a new editor window and view them side by side instead of having to jump up and down in the same file
Yeah, I know that some editors have a "split window" mode, but those don't play nicely with multiple monitors, etc, etc
I'm making a simple game with SFML 1.6 in C++. Of course, I have a lot of picture, level, and data files. Problem is, I don't want these files visible. Right now they're just plain picture files in a res/ subdirectory, and I want to either conceal them or encrypt them. Is it possible to put the raw data from the files into a resource file or something? Any solution is okay to me, I just don't want the files exposed to the user.
EDIT
Cross platform solutions best, but if they don't exist, that's okay, I'm working on windows. But I don't really want to use a library if it's not needed.
Most environments come with a resource compiler that converts images/icons/etc into string data and includes them in the source.
Another common technique is to copy them into the end of the final .exe as the last part of the build process. Then at run time, open the .exe as a file and read the data from some determined offset, see Embedding a filesystem in an executable?
The ideal way for this is to make your own archive format, which would contain all of your files' data along with some extra info needed to split files distinctly within it.
This might be difficult to answer. I am trying to understand some codebase and in one of the project folders named FOO there is a main.cpp file plus other .hpp and .cpp files. These filenames are made descriptive using underscores such as Hi_there_file.cpp etc.
There are two files Foo.tab.hpp and Foo.tab.cpp
Is *.tab.cpp an indication of something or just a poor try at making the file name descriptive?
Turns out, that file is generated by a perl script so they put a tab.cpp at the end. maybe even delete this question. Very very specific to what i'm working with