Difference between revisions of "DIP76"

From D Wiki
Jump to: navigation, search
(Initial edit)
 
m (formatting)
Line 33: Line 33:
 
== Benefits ==
 
== Benefits ==
  
* Becomes nothrow, which enables code that uses it to become nothrow.
+
* Becomes nothrow, which enables code that uses it to become nothrow. Nothrow code does not need to produce exception unwinding wrappers, which
Nothrow code does not need to produce exception unwinding wrappers, which
 
 
can improve the performance of using ref counted objects.
 
can improve the performance of using ref counted objects.
 
* Becomes @nogc, because the thrown exception is allocated on the GC.
 
* Becomes @nogc, because the thrown exception is allocated on the GC.
 
* Becomes consistent with the behavior of std.utf.byDchar().
 
* Becomes consistent with the behavior of std.utf.byDchar().
 
* Will be somewhat faster.
 
* Will be somewhat faster.
* Current industry practice has evolved towards using replacement characters
+
* Current industry practice has evolved towards using replacement characters for bad UTF rather than failing.
for bad UTF rather than failing.
 
 
 
  
 
== Breaking Existing Code ==
 
== Breaking Existing Code ==

Revision as of 03:16, 7 April 2015

Title: Autodecode Should Not Throw
DIP: 76
Version: 1
Status: Draft
Created: 2015-04-06
Last Modified: 2015-04-7
Author: Andrei Alexandrescu and Walter Bright
Links:


Abstract

When the autodecoder encounters invalid UTF, it throws an exception. This proposal changes that so it replaces the invalid UTF with the Replacement Character.

Benefits

  • Becomes nothrow, which enables code that uses it to become nothrow. Nothrow code does not need to produce exception unwinding wrappers, which

can improve the performance of using ref counted objects.

  • Becomes @nogc, because the thrown exception is allocated on the GC.
  • Becomes consistent with the behavior of std.utf.byDchar().
  • Will be somewhat faster.
  • Current industry practice has evolved towards using replacement characters for bad UTF rather than failing.

Breaking Existing Code

This can break existing code, although it seems highly unlikely that any programs actually require the decoding to throw on invalid UTF.


Deprecation Path

There doesn't seem to be any deprecation path for this change. It will have to be abrupt.

NG Announcement

Copyright

This document has been placed in the Public Domain.