get a process by name and get a processes' descendants

[giampaolo]

From [email protected] on January 19, 2012 11:39:21

    Here is a diff for two new additions to psutil.
    One is psutil.proc_by_name, which g.rodola actually wrote in issue 241 . I just 
    included it here in the diff, and changed it from returning an iterator to a list.

    The other is a get_lowest_descendants method on the Process object, which 
    fetches all descendants from a process, that don't have child processes 
    themselves.

    The use case that I have for this method is that I want to see what I am 
    currently doing in my terminal application. So my terminal will have a couple 
    of bash child processes (one for each tab), and each of those has child 
    processes as well, some being an ssh command into another machine, some being a 
    python interpreter in a certain folder, or a vi instance on a certain file, etc.

    With the get_lowest_descendants method I can easily fetch those descendants and 
    what they are doing.

Original issue: http://code.google.com/p/psutil/issues/detail?id=242

[giampaolo]

From [email protected] on January 19, 2012 02:39:37

and now with the diff attached :)

Attachment: init.py.diff

[giampaolo]

From g.rodola on January 19, 2012 05:33:39

> patch in attachment adds a get_lowest_descendants()
> method on the Process object, which fetches all 
> descendants from a process, that don't have 
> child processes themselves.

Mmm interesting.
I think I don't have a clear mind about this yet.
At first it seems kind of useful, on the other hand I'm not sure about its 
reliability.
Imagine this:

process A creates process B which creates process C

In normal circumstances your method would return A, B and C, but if B gets 
terminated in meantime you're gonna lose the reference to C and get only A in return.

Other than that, I think this would be better served by adding an extra 
argument to get_children() method:

>>> get_children(recursive=True)

...and make it return only B and C but not itself (A).


Also, it seems natural to make get_children() return an iterator rather than a list.
The same should be done for get_open_files(), get_connections() and 
get_threads() (I'll file a new ticket for that).


Jay, what do you think?

Status: WaitingForReview
Cc: [email protected]

[giampaolo]

From [email protected] on January 19, 2012 06:07:13

In your example where process A creates process B which creates process C, 
the code would only return process C, not B, since B has children.

Another option for my use case would then be 
>>> filter(lambda x: len(x.children)==0,get_children(recursive=True))

[giampaolo]

From g.rodola on April 07, 2012 08:38:10

This is now implemented in r1282 .
Please look at the docstring and let me know if with this you're able to cover 
your use case.

Status: Started

[giampaolo]

From g.rodola on April 08, 2012 05:37:10

Change recursion algorithm as r1283 .

[giampaolo]

From g.rodola on April 12, 2012 11:15:08

Status: FixedInSVN
Cc: [email protected]
Labels: Milestone-0.5.0 Usability

[giampaolo]

From g.rodola on June 27, 2012 11:54:03

0.5.0 is finally out. Closing out as fixed.

Status: Fixed

[giampaolo]

From g.rodola on March 02, 2013 04:06:27

Updated csets after the SVN -> Mercurial migration: r1282 == revision 
19893ece7eff r1283 == revision 0e9fabea61e2