for / while else doesn't make sense

Rustom Mody rustompmody at gmail.com
Wed Jun 15 13:03:41 EDT 2016


On Wednesday, June 15, 2016 at 9:24:21 PM UTC+5:30, Random832 wrote:
> On Wed, Jun 15, 2016, at 10:20, Rustom Mody wrote:
> > Claim is that the damaging propensities of goto are replicable with
> > break.
> 
> The "damaging propensity" in this particular case simply comes from the
> fact that it's a statement that can appear in the body of an if
> statement (the real flaw that caused the bug here here was the optional
> braces of the if statement), which is just as true of anything else. It
> could have called a function that it shouldn't have called.

People have adduced various reasons:
1. Optionality of braces
2. gcc silently non-mentioning dead code
3. indentation and braces mismatch -- were it python(ish) the second goto would
be harmless dead code INSIDE the if
4. Careless merging
5. Culture of cut-paste
6. All kinds of soft questions -- code should be reviewed/pair-programmed/etc
7. Even more irrelevant soft point -- Apple is a bad company for being thus irresponsible
8. etc 

> 
> But a lot of people do claim that break should be generally "considered
> harmful" just like goto and never used at all because of this
> isomorphism.

Minor point: I dont think isomorphism is correct
As I said the goto can originate anywhere and target to anywhere (in function scope). Not break
longjmp can scale function scopes but has other restrictions like respecting the
stack, the setjmp-site needs to have been visited first.
IOW different unstructured constructs have differing properties

Major point:  "Considered harmful" should be considered harmful :-)
IOW cargo-cult programming without any understanding is worse than specific fads
There are things -- eg automata -- where the most faithful rendering is with
gotos: : labels for states, gotos for transitions.  Any 'more structured' solution is usually more messy.

OTOH Duff's device shows terrifyingly unstructured code with nary a goto/break 
in sight

So break also has its uses. Lets just not pretend its structured


More information about the Python-list mailing list