"mprof run <executable>": What does each output column represent?

memory-profiling, python

Solution

1st column: "MEM" is just a label.

2nd column: RSS (resident set size) in megabytes. So your process ended up using 8.8 MB.

3rd column: Unix time stamp:

>>> import time
>>> print(time.ctime(1424769615.1960))
Tue Feb 24 01:20:15 2015

It looks like `mprof` just logs the process size over time. If you want more specific details about the internals of your program (line-by-line) it looks like you can use `mprof run --python file.py` to get that information (you also need to decorate your functions with `@profile`, see the Usage section in the documentation).

Edit: In response to the ImportError & traceback you posted: You are getting that error because you're running the file as `"../myfile.py"`. Since you're outside of the `bsite` directory it can't find the `config.py` file.

You need to run the program like this:

 $ cd /home/vikas/bsite/
 $ /home/vikask/memory_profiler-0.32/mprof run --python myfile.py

This way you're in the correct directory for the `import` to work.

Another way to get it to work is to build your project as a python package. When you do this your files can be properly imported from anywhere. This is a little complicated to get set up, but it's a pretty good idea.

The other more hacky option is to modify `PYTHONPATH` or `sys.path` to get the imports to work. For example like this:

$ cd /home/vikask/memory_profiler-0.32/
$ PYTHONPATH=/home/vikas/bsite/ ./mprof run --python ../myfile.py

Finally, it looks like you're running the `mprof` command directly from the memory_profiler source. You can install the package by doing "`python setup.py install`" in that directory (or by using `pip install memory_profiler`). Then you should have the `mprof` command on your PATH and can run it from any directory.

Problem

Original source