On Tuesday, 22 July 2014 at 23:43:59 UTC, Robert burner Schadek
wrote:
On Tuesday, 22 July 2014 at 21:52:09 UTC, linkrope wrote:
Controversial conditional logging
----------------------------
The only advantage of
tracec(condition, "passed");
over
if (condition) trace("passed");
would be, that a costly evaluation of the condition is omitted
when there is no trace logger.
That's why the std.log proposal had 'when(lazy bool now)'.
First, I was puzzled about your argument that LOG_FIRST_N or
LOG_EVERY_N would be no problem with the '...c' functions. But
a look at the implementation confirmed that the std.logger has
no lazy evaluation of the condition; discarding the only
advantage.
passing a bool as a delegate that is the first thing that gets
evaluated does not sound right.
Indeed: that's why the lazy condition should be evaluated last!
Your "will log" condition is very simple and efficient: only >=
and !=
But you cannot know, how long the evaluation of the user-provided
condition will take.
BTW: with 'globalLogLevel', 'defaultLogger.logLevel'and
'LogLevel.off' the usual 'willLog' predicate will come in handy
to avoid code duplication.
While the lazy evaluation of the condition would be the only
advantage over
if (condtion) log(...);
I haven't encountered a single opportunity for conditional
logging in the code of our company.
'if'/'log' is always followed by 'return', 'break', 'continue',
...
if (condition)
{
log("will do something else because condition passed");
return;
}
