Re: a new language, designed for safety !

["Dan'l Miller" <optikos@verizon.net>]

On Monday, June 2, 2014 8:37:12 PM UTC-5, Nasser M. Abbasi wrote:
> Fyi, for those who like safe languages, watch out Ada and
> 
> Spark, Apple introduced new "safe" language today, meant
> to replace objective-C:
> 
> "In addition, Apple notes how the language was designed for
> safety, with variables that have to be initialized before
> use <WOW!>, arrays and integers that are checked for
> overflow <WOW!> and automatic memory management."

One major category of Objective-C unsafety that Swift apparently does absolutely nothing to prevent (until fatal error at runtime, hence the unsafety) is whether an object actually provides a method for a given selector or whether an object actually provides a forwarding to another object for a given selector or is handled by a per-class or per-object-instance error handler or none of the above, which precipitates the fatal error at runtime.

Except when an iPhone or iPad or iPod or MacOSX app presents to the human user a roulette wheel of all object instances or of all class names to choose an object instance or class to extend dynamically at runtime, the programmer certainly knows at source-code-editing time which categories of object instances or which  class name is to have additional methods or forwardings for a given selector--and hence when to create some overtly-declared-at-compile-time categorization (e.g., subclass) of a class or of a categorization of object-instance.  Hence, the vast vast vast majority of so-called dynamic extension of an object instance or of a class is actually an engineering-time decision (i.e., prior to compile-time).  Therefore, Objective-C and Swift both provide no compile-time-assured mechanism to 100% assure that a given class (or a given object instance) will able to successfully invoke a method or forwarding when "sent" a "message" for a given selector.  This is not surprising for the 1980s-vintage Objective-C, back when Smalltalk was often held up as the ultimate state-of-the-art in OO; Smalltalk had much the same fundamental unsafety, due in turn much the same selector mechanisms as Objective-C.  But almost 3 decades later, Swift has the same fundamental unsafety at its core and throughout all nontrivial source code, when following Apple's advice regarding when so-called dynamic extension of a class or of a object-instance is to be utilized.

Or in short, Swift's so-called safety is much more like swiss cheese (Swift cheese?) than Ada's higher expectations of what constitutes safety.