[Date Prev][Date Next][Thread Prev][Thread Next]
[Date Index]
[Thread Index]
- Subject: Re: Lua 5.4.0-alpha-rc1 behavior with closing methods errors
- From: Francisco Olarte <folarte@...>
- Date: Sat, 22 Jun 2019 16:17:55 +0200
David:
On Sat, Jun 22, 2019 at 11:20 AM David Heiko Kolf <david@dkolf.de> wrote:
> Francisco Olarte wrote:
> > I'm nearly sure you are mixing concepts here, comparing try/finally
...
> Unfortunately I haven't worked with Java for a long time but in C#
> "using" and "finally" are actually equal and not different animals:
> <https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/keywords/using-statement>:
> > You can achieve the same result by putting the object inside a try
> > block and then calling Dispose in a finally block; in fact, this is
> > how the using statement is translated by the compiler.
Well, the same thing can be said of Java, and of many other languages
which have with/using and try+finally. Using is just a shortcut which
writes some code for you, but the same can be said of many constructs.
What I was trying to bring forward is that when you make a program you
use with/using for objects which have a purposefully built method,
which does not throw errors, and which is / has benn tested
separatelly. You use finally when your clean up block is more
convoluted and needs logic depending on local state or similar things.
> So far I have only used finally (in any language) for cleaning up
> resources, too.
But you do not use to clean up resources which are prepared to clean
up themselves and tested, except when you are in one of does languages
which force you to do it ( like java pre 7 ). Normally, when you write
a library class, say a socket class, you add some closing method which
can return errors, or throw, and may need the object alive to be
processed. Then you add a method which does some "standard" error
handling on the low-level one but wwhich does not throw and can be
called without worrying about the result. I.e., closing a writable
file can, and does, fail with ENOSPC. Closing a socket may fail if the
other side is down. Some clients may need the details, but normally
you just close it and forget. And, if your language supports
with/using constructs, you normally enable the class to be used with
them. But you do not normally use the low level, exception throwing,
variant there, as it leads to difficult to read code. If you need this
you use try/finally and call the more detailed close and handle the
errors explicitly there. There are exceptions, but normally languages
which allow faults in the closing tend to supress it and ignore the
error, as a well written close method normally just throws in extreme
cases.
> Maybe my first example and some of my words weren't precise enough, but
> anyway, what I meant is I want to be informed in case a cleanup method
> for a resource didn't finish its job.
That is part of whatI want to establish. A properly written cleanup
method, designed for cleanup, should always finish the job. If your
resource cannot be properly finalized, you should not use an
automatically-called cleanup method, but write a finally block where
the handling is explicit. I've found trying to do it the other way
leads to lot of pain. Not all resources lend themselves to automatic
cleanup.
Francisco Olarte.