[Issue 948] New: operatoroverloading.html - Rationale section is both out of date and incomplete
d-bugmail at puremagic.com
d-bugmail at puremagic.com
Sun Feb 11 16:00:59 PST 2007
http://d.puremagic.com/issues/show_bug.cgi?id=948
Summary: operatoroverloading.html - Rationale section is both out
of date and incomplete
Product: D
Version: 1.005
Platform: All
URL: http://www.digitalmars.com/d/operatoroverloading.html
OS/Version: All
Status: NEW
Severity: normal
Priority: P2
Component: www.digitalmars.com
AssignedTo: bugzilla at digitalmars.com
ReportedBy: smjg at iname.com
At the end of the "Overloading <, <=, > and >=" section is a "Rationale", which
states:
"For some objects, testing for less or greater makes no sense. For these,
override opCmp() with: " [...]
There are two things wrong with this (besides the fact that that code would now
trigger a "statement is not reachable" warning in DMD, but that's irrelevant):
- The instruction to override opCmp is a leftover from when the default
Object.opCmp compared the memory address. Nowadays, Object.opCmp already
throws an error, so there's no point overriding a class's opCmp as an extra.
- This has been briefly mentioned before but nothing done about it. On reading
just above it that "every class object has a opCmp()", people will look for an
explanation of _why_ opCmp should be defined for every class. Instead, the
average reader is treated to what reads as an argument _against_ this, namely
"For some objects, testing for less or greater makes no sense."
I propose the following rewrite of the bullet points in this section to address
both problems:
- Testing for equality can sometimes be a much more efficient operation than
testing for less or greater than.
- Having opCmp defined in Object makes it easier to implement associative
arrays and sorting efficiently.
- For some classes, testing for less or greater makes no sense. The default
Object.opCmp throws an error at runtime to indicate that the programmer has not
defined an opCmp for the class being used.
--
More information about the Digitalmars-d-bugs
mailing list