does lack of type declarations make Python unsafe?

Bruno Desthuilliers bdesth.nospam at removeme.free.fr
Tue Jun 17 16:00:37 CEST 2003


Tim Rowe wrote:
> On Mon, 16 Jun 2003 21:01:42 +0000, Bruno Desthuilliers
> <bdesth.nospam at removeme.free.fr> wrote:
> 
> 
>>Tim Rowe wrote:
>>(snip)
>>
>>
>>>So if we're going to have meaningless
>>>polemic, let me try mine :-)
>>
> 
>>Please be serious. 
> 
> 
> I consider it at least as serious as the post I was replying to,
> though in my case the exaggeration to the point of comedy was
> intention (I don't know about successful).  No, the compiler and other
> static testing won't catch all the bugs.  But it has been pretty
> thoroughly shown that dynamic testing won't catch them all either.

I dont pretend that any known methodology can catch *all* bugs.

> And the research I've seen
Links, please...

> indicates that the two techniques catch
> pretty much non-overlapping sets of bugs.

Well... With static-typing and compilation, the compiler can catch type 
bugs. They are not the most common bugs, nor the more difficult to catch  .

> Go back to my parody of Anton's position: which do you think we can do
> without in our diet? Fibre or vitamins?  Ok, now which do you think we
> can do without in our software assurance?  Static testing or dynamic
> testing?

That comparison does not stand. We need both fiber and vitamins to live, 
but the fact is that programs written with dynamic languages works, and 
are no more no less buggy than programs written with static languages.

> (As an aside, when the project manager comes along and says we have to
> cut something to meet timescales, do you think it will be compilation
> or test that's more likely to be cut?)

Which compilation ? Remember this is c.l.python here... No overnight 
compilation phase needed.

Now, if the project manager is stupid enough (yes, don't tell me, many 
of them are...) to cut on the test, you're in the same trouble, 'static 
testing' or not. The fact that a C++ source compile doesn't means it's 
bug-free, AFAIK...

Bruno Desthuilliers





More information about the Python-list mailing list