TweetFollow Us on Twitter

September 90 - Assertions in MacApp

Assertions in MacApp

James Plamondon

Have you ever loved someone you couldn't have? Someone of such rare grace, beauty, and elegance that his or her charms were beyond the appreciation of many of your peers, whose tastes, shall we say, were not nearly as refined and discriminating as your own?

Such is my love affair with Bertrand Meyer's object-oriented language, Eiffel. Well described in his book, Object-oriented Software Construction (Prentice-Hall, 1988), Eiffel is a paragon of elegance, power, and simplicity. Every time I waste an hour fighting Pascal's USES clause, or stuff some routine into UObject to get around single inheritance, or write yet another list class (because TList only handles object references), or consider (out of desperation, not desire) turning to C++, I long for the simple solutions Eiffel would provide.

But my love is as hopeless as it is deep-there is, at the time of this writing, no implementation of Eiffel available for the Mac. There's a stripped-down A/UX version nearing completion, and a full Mac OS version planned for a January release-however, these are still in vapor form. Until they solidify-and mature into the complete implementations they almost certainly will not be, in their initial releases-my love must remain unrequited, its fires banked, but still glowing.

Until then, I will continue to console myself with Object Pascal. There are three reasons why I do so:

  • Object Pascal has a sterling virtue that even Eiffel can't claim: it's the lingua franca of MacApp.
  • If you can't be with the one you love, baby… love the one you're with!
  • MacApp's comprehensive utilities mask some of Object Pascal's deficiencies such that, if the bar's dark, the beer's cheap, and you squint a little bit, Object Pascal can look just enough like Eiffel to get you interested.

I'm going to address one aspect of that resemblance in this annotated compilation of MacApp.Tech$ links: assertions. An integral part of Eiffel, assertions are the means by which the semantics of an abstract data type are enforced. For example, Eiffel's preconditions (one kind of assertion) ensure that a method is never called with invalid arguments. Likewise, postconditions ensure that a routine has always done what it was supposed to have done before it returns. Loop variants and invariants ensure loop termination. While not allowing the direct production of formal, mathematical proofs directly from code, the Eiffel assertion mechanism is a powerful weapon in the commercial programmer's arsenal against bugs.

Neither Object Pascal nor MacApp directly address these issues. However, hidden away in the purdah of UMacAppUtilities, unmentioned, undocumented, and generally ignored, are the seeds of an assertion mechanism: the Assertion() global routine, and the gPrecondition global constant/variable. The use of these features are the topic of the discussion below.

Finally, I'm going to start this compilation out of sequence, with my response to a link from Mr. Cooley of After Hours Software. His link (which is, unfortunately, missing from the archives) requested an explanation of what assertions are and what they're good for. My response was actually the end of this thread, but in a zen-like manner not uncommon in programming, the end is really the beginning.

Please note that italicized comments in brackets [like these] have been added by me for the purpose of annotation.

5/2/90
POWERUP.ENG Power Up Software,PRT
Dear Mr. Cooley,

Pretty much everything I know about assertions I learned from Bertrand Meyer's book, Object-Oriented Software Construction, published in 1988 by Prentice-Hall.

The first four chapters of this book are the best discussion I've found anywhere of what makes some software "good" and other software "bad," and how OOP facilitates the production of "good" software.

A lot of the rest of the book is a discussion of other OOP issues - multiple inheritance, genericity (aka parameterized types), garbage collection, classes vs. types, exception handling, assertions, etc.. These issues are discussed in the context of Meyer's language, Eiffel, which I think is the greatest programming language ever, because of its elegant encapsulation of all of the ideas presented in the book. [Hear, hear.] (A number of companies are currently planning to produce Eiffel for the Mac; it already is available for A/UX.)

Assertions are statements of fact that are checked at runtime. If the asserted boolean condition evaluates to FALSE, then a run-time exception is raised - essentially a break into the debugger.

Assertions fall into three rough categories: class invariants, pre- and post- conditions, and other assertions.

The class invariant is checked after every method call, to ensure that the basic characteristics of the class are maintained (that an empty stack contains no elements, for example).

A method's preconditions are checked before entry into the method, to make sure that the method is being called in a valid fashion. For example, one cannot meaningfully pop from an empty stack, so the Pop() method would have a precondition stating that the stack was not empty. If it were empty when Pop() was called, then the assertion would be false, and an exception would be raised.

Likewise, a postcondition ensures that the method does what it is supposed to do. Pop() would include a postcondition stating that, after the method was done, the stack would contain one fewer element than before Pop() was called.

Other assertions include loop assertions (variants and invariants) and the check assertion, which is like the Assertion() routine in MacApp.

All of the assertions are present to ensure that the code does what it is supposed to do. They also enhance the code readability and maintainability, by making it crystal clear what each method is supposed to do, and under what circumstances it can be expected to do it.

That's a quick guide to assertions in Eiffel. They existed in other languages before Eiffel - as a number of people have pointed out to me - but the Eiffel implementation is pretty slick, nonetheless. For a more detailed discussion, I'd definitely see Meyer's book.

Looking forward to Mac Eiffel (later this year? Early in '91? That pesky memory manager sure causes problems!),

James Plamondon

[With that overview of assertions out of the way, we can return to the beginning of the thread.]

4/19/9
POWERUP.ENG Power Up Software,PRT
Dear Steve,

I was wandering around in MacApp 2.0B12 today, and I stumbled across a few interesting things that made me think that somebody has been taking a close look at Eiffel.

First, in UFailure, I found an the following procedure:

Assertion(condition: Boolean; description: StringPtr);

which is only ever called in TApplication.Idle(), as follows:

Assertion(gTarget <> NIL, AtStr('gTarget <> nil'));

Then, wandering further, I encountered two variables (or constants, depending on the state of qDebug): gPreCondition and gPostCondition, which are defined in UMacAppUtilities.p and initialized (to TRUE) in DoRealInitToolBox(), which is in UMacAppUilities.inc1.p. These entities are never referred to anywhere else in MacApp.

These look awfully useful - but the fact that they are not used anywhere in MacApp suggests that they are not safe to use in one's application. So, what's up?

James Plamondon

4/19/9
A14 Carnegie Mellon, Rob Chandok,PRA

James writes: I was wandering around in MacApp 2.0B12 today, and I stumbled across a few interesting things that made me think that somebody has been taking a close look at Eiffel.

I find it interesting that people think Eiffel invented assertions programming languages. While Eiffel is "way cool," it isn't the first language to think about assertions.

I find it more useful to think about good features in terms of the abstract concept, since that is the real importance. You can do assert(boolean,string) in any language. We have such a built-in procedure in our Pascal teaching environment. But we didn't get it from Eiffel.

There has been other interesting work in programming languages before Eiffel, although Eiffel is a well thought out language designed by a smart person. Besides, real assertions would and some way of saying "there exists" and "for all." :-)

Rob

4/19/9
ALGER Alger, Jeff,VCA
Rob,

Amen. There is too much "discovery" of old ideas out there, and assertion is one of them. All major languages are Turing machines, so there is no difference between what you can, in theory, do from one language to another. The differences are all in degree of difficulty to accomplish certain objectives versus execution time and space.

On the subject of "for all" and "there exists," that (sigh) would be a welcome addition to Object Pascal, but requires a good deal more infrastructure than OP provides. Specifically, OP does not have the concept of a set of objects, let alone placing all objects into "master" sets based on class or anything else.

I liked the term "syntactic sugar" applied to assertions in Eiffel, but would take that one step further. Objects in OP are, to me, syntactic sugar plus a little bit of dispatch glue, not objects in the same sense as Smalltalk, CLOS, Object Prolog, etc., etc., etc. "Even" C++ allows allocation and deallocation of storage of objects to be determined separately from the behavior of the objects themselves (a class behavior), class variables, and a number of other blue-blood OOP features. In C++, one can at least conveniently superimpose maintenance of class sets and, therefore, "for all" and "there exists," without changing the public interface to a class's objects. C++ has its problems, too, but the point here is that one can and should do MUCH better than OP.

Ah, nothing like stirring up a little controversy in the morning.

Jeff Alger, Exis

4/19/9
V0683 Amoco Tech, Eric Berdahl,VAR

Yes, the Assertion mechanism is nice to have, especially since OPascal doesn't have language support for the mechanism (I actually think that Eiffel's construct is more syntactic sugar than anything else).

[Every expression we write, unless we're writing in machine language, is "syntactic sugar." Implementing assertions at the library level detracts from their utility (compared to implementing them at the language level), even if it can be done. As I tell my son at the dinner table, if you mix your food together on the plate, it loses something, even if it's going to get all mixed up by the time it gets to your stomach!]

If they [assertions] aren't safe to use, you couldn't prove it by me. I've been using assertions for a while with no consequences. Remember, though, that assertions are things that really are conceived of and placed during object design. If they were added anytime after a significant amount of MacApp design took place, they may not be very easily placed.

[Virginity, once lost, is rarely restored.]

Eric Berdahl

4/20/9
POWERUP.ENG Power Up Software,PRT
Dear Rob, Jeff, and Gentlepersons Everywhere,

Touché! You are correct both in that assertions are not new to Eiffel, and that "there is too much re-discovery of old ideas out there." I first encountered language-based assertions in Eiffel, though, and was not aware that they had been supported in any earlier languages. Thank you for your correction.

I am reminded of a girl whom I once overheard asking a friend, "Did you know that Paul McCartney was in a band before Wings?" [Yes, it's a true story, I swear!] [Ed: I must be slow or something. What band?]

James Plamondon

4/19/9
MACAPP.TEST MacApp SQA Team
James,

As one of the people trying to figure out how you ensure the correctness of a product like MacApp, I am intrigued by Eiffel's assertion mechanism. The routine you found (which was there in B9) is a small step, but it's worth noting that since Eiffel has language support for this stuff, they can add things like the "old" and "nochange" notations (p.115 of the good book), as in

ensure
Size = old fSize + 1

Just because MacApp doesn't use a routine doesn't mean you shouldn't (OK, so we forgot the VAR parameter in GetPortTextStyle and didn't catch it because we didn't use it, but anyway...). [So that's why one of my program's features didn't work! I just wrote it yesterday evening, using GetPortTextStyle, and I was wondering what I was doing wrong!]

Please share your suggestions for assertions with us. Do you think we should have a boolean function TObject.ClassInvariant? Homework: Write preconditions and postconditions for TApplication.TrackMouse.

Richard Rodseth, MacApp Quality

4/20/9
POWERUP.ENG Power Up Software,PRT
Dear Richard, et al.,

I would not know where to begin to write preconditions and postconditions for TApplication.TrackMouse()!

Assertions may work best for container classes (stacks, lists, tables, etc.), which have a mathematically definable nature. TList looks like the best place to start using it-especially given the wide use of lists in MacApp.

Now that I know that there are no impediments to the use of the routine Assertion(), I may very well start using Assertion() in my own code. […and so I have; the little buggers are sprinkled though my code like salt on popcorn.]

Of course, it would help if one could change the values of gPreCondition and gPostCondition from the Debug menu… but I'll leave that as an exercise for the reader. [Hint: use the UDebug routine DebugFlag().]

As to TObject.ClassInvariant(), it sounds like a great idea. Perhaps the following suggestion is possible:

  • Add the field fCheckInvariant (a boolean) to TObject (but only if qDebug is true).
  • Allow the user to set the state of TClassName.fCheckInvariant for each class, independently, in the MacApp debugger.
  • If TClassName.fCheckInvariant were true, then the MacApp debugger would call TClassName.ClassInvariant() (polymorphically, of course) after each call to a method of TClassName (other than ClassInvariant()). The MacApp programmer would NOT have to call TClassName.ClassInvariant() herself; that would defeat the purpose.

[Actually, the CheckInvariant flags would probably be better implemented using a table of boolean values, indexed by class ID.]

This might be exceedingly slow, but it would also be exceedingly nifty. The programmer would, of course, have to write her own ClassInvariant() overrides. Likewise, any precondition and postcondition assertions would have to be written by the programmer, and included in her code, since they would vary from method to method.

Now, if we can just add multiple inheritance and genericity to MacApp (while shovelling in some added simplicity, as Steve Friedrich would say), we can make MacApp and Object Pascal a real working system.

Looking forward to a Mac version of ANY LANGUAGE WHICH SUPPORTS ALL THE THINGS I LIKE ABOUT EIFFEL (but not necessarily Eiffel itself-one must keep an open mind, after all), I remain

James Plamondon

4/20/9
A14 Carnegie Mellon, Rob Chandok,PRA

Rich Rodseth's notes about corrrectness are interesting. Here are some more points about assertions.

While Eiffel can do things like

Size = old fSize + 1

doesn't it seem odd that you have to specify that and the code to add 1 to fSize? The difficulty to me is that you don't want two separate representations of your program–specification and implementation. In this case, you have to make sure that both your code and your invariant is correct. Seems like you might introduce bugs in the invariants as easily as in the code.

[…and that's a very good point. I have indeed written incorrect assertions and wondered why the heck they were (a) flagging errors that weren't errors (easily located and fixed), or (b) not flagging real errors (a much more subtle and insidious problem). I am open to suggestion on ways around this, which aren't themselves error-prone.]

But this is the real world, I know. I haven't seen any language that handle this, although I know of some promising research directions.

In addition, the idea of TObject.ClassInvariant is founded in the premise that inheritance requires the subclass to maintain the behavior of the superclass. This is contrary to many uses of inheritance. Let's see if I can rattle off some ideas off the top of my head.

There are two "kinds" of inheritance: protocols and base class. If you inherit in the protocol style, you are basically saying that "subclass S is just another kind of superclass T, with different internal details, and if you send message M to S, it will behave within the definition of T." I can fully define what the "sort" message will do, no matter what subclass of a sortable class I send the message to.

That doesn't jive with TApplication.TrackMouse, which is an example of inheriting in the "base class" style. No one could possible know what the message TrackMouse is going to do for any subclass of TApplication.

Bruce Horn and I have spent much time arguing over these issues. In addition, you might think about inheritance in this way: If you want to enforce pre/post conditions in the superclass, should the superclass control when the OVERRIDDEN methods get called? For example, just because you override DRAW, should that prevent your superclass's DRAW from being called ? What if DRAW is responsible for maintaining the superclass invariant? Imagine a world where a method call could either percolate up via calls to INHERITED (like object pascal) or down from the superclasses (which is similar to wrappers).

Well, that's enough for now. Suffice to say that invariants are tough. We can barely prove tiny programs correct, it's no wonder we don't know how to specify a language even for executing assertions.

[This is certainly true. Meyer has taken a lot of heat about Eiffel's assertions not being a "complete" formal system for automatic generation of proofs of correctness. But it was never intended nor presented to be such a mechanism-rather, it's just a means by which more robust code can be written. They're not perfect, but they're a big step in the right direction-and if you're dying of thirst, even muddy water looks good!]

I remain, as perplexed as ever,
Rob

Annotator's Summary

The UFailure routine Assertion() is there to be used. It is a powerful and convenient means of ensuring that some condition is true at runtime. I use it frequently to check handles vs. NIL, check values for correctness, and anything else that can be summed up in a BOOLEAN statement. If the condition is FALSE at runtime, the program drops into the MacApp debugger, and asks if you want to signal failure. If included in non-debug code, an Assertion() failure signals a silent failure. Using assertions liberally has saved me hours-days, by now-of debugging time.

The Boolean gPrecondition is a global variable (initialized to TRUE) if qDebug is TRUE, else it is a global constant (with the value FALSE). By adding a call to DebugFlag() in your application's initialization routine, you can toggle this flag on and off in the MacApp debugger. To use it, place the assertions which check a routine's arguments for validity inside an IF gPrecondition statement. In qDebug mode, these assertions will only be checked if the variable gPrecondition (as set by you in the debugger) is TRUE. In non-qDebug mode, gPrecondition is a constant with the value FALSE, so the assertions are stripped out by the linker.

Eiffel is not a perfect language; even blinded by infatuation as I am, I can see that. But Object Pascal-whose charms have long since begun to fade-could do far worse than to imitate Eiffel's standard of beauty.

[Ed: James can be contacted (for what would no doubt be a lengthy discussion] at AppleLink address POWERUP.ENG.]

 
AAPL
$111.78
Apple Inc.
-0.87
MSFT
$47.66
Microsoft Corpora
+0.14
GOOG
$516.35
Google Inc.
+5.25

MacTech Search:
Community Search:

Software Updates via MacUpdate

NeoOffice 2014.6 - Mac-tailored, OpenOff...
NeoOffice is a complete office suite for OS X. With NeoOffice, users can view, edit, and save OpenOffice documents, PDF files, and most Microsoft Word, Excel, and PowerPoint documents. NeoOffice 3.x... Read more
LibreOffice 4.3.5.2 - Free Open Source o...
LibreOffice is an office suite (word processor, spreadsheet, presentations, drawing tool) compatible with other major office suites. The Document Foundation is coordinating development and... Read more
CleanApp 5.0.0 Beta 5 - Application dein...
CleanApp is an application deinstaller and archiver.... Your hard drive gets fuller day by day, but do you know why? CleanApp 5 provides you with insights how to reclaim disk space. There are... Read more
Monolingual 1.6.2 - Remove unwanted OS X...
Monolingual is a program for removing unnecesary language resources from OS X, in order to reclaim several hundred megabytes of disk space. It requires a 64-bit capable Intel-based Mac and at least... Read more
NetShade 6.1 - Browse privately using an...
NetShade is an Internet security tool that conceals your IP address on the web. NetShade routes your Web connection through either a public anonymous proxy server, or one of NetShade's own dedicated... Read more
calibre 2.13 - Complete e-library manage...
Calibre is a complete e-book library manager. Organize your collection, convert your books to multiple formats, and sync with all of your devices. Let Calibre be your multi-tasking digital librarian... Read more
Mellel 3.3.7 - Powerful word processor w...
Mellel is the leading word processor for OS X and has been widely considered the industry standard since its inception. Mellel focuses on writers and scholars for technical writing and multilingual... Read more
ScreenFlow 5.0.1 - Create screen recordi...
Save 10% with the exclusive MacUpdate coupon code: AFMacUpdate10 Buy now! ScreenFlow is powerful, easy-to-use screencasting software for the Mac. With ScreenFlow you can record the contents of your... Read more
Simon 4.0 - Monitor changes and crashes...
Simon monitors websites and alerts you of crashes and changes. Select pages to monitor, choose your alert options, and customize your settings. Simon does the rest. Keep a watchful eye on your... Read more
BBEdit 11.0.2 - Powerful text and HTML e...
BBEdit is the leading professional HTML and text editor for the Mac. Specifically crafted in response to the needs of Web authors and software developers, this award-winning product provides a... Read more

Latest Forum Discussions

See All

Make your own Tribez Figures (and More)...
Make your own Tribez Figures (and More) with Toyze Posted by Jessica Fisher on December 19th, 2014 [ permalink ] Universal App - Designed for iPhone and iPad | Read more »
So Many Holiday iOS Sales Oh My Goodness...
The holiday season is in full-swing, which means a whole lot of iOS apps and games are going on sale. A bunch already have, in fact. Naturally this means we’re putting together a hand-picked list of the best discounts and sales we can find in order... | Read more »
It’s Bird vs. Bird in the New PvP Mode f...
It’s Bird vs. Bird in the New PvP Mode for Angry Birds Epic Posted by Jessica Fisher on December 19th, 2014 [ permalink ] Universal App - Designed for iPhone and iPad | Read more »
Telltale Games and Mojang Announce Minec...
Telltale Games and Mojang Announce Minecraft: Story Mode – A Telltale Games Series Posted by Jessica Fisher on December 19th, 2014 [ permalink ] | Read more »
WarChest and Splash Damage Annouce Their...
WarChest and Splash Damage Annouce Their New Game: Tempo Posted by Jessica Fisher on December 19th, 2014 [ permalink ] WarChest Ltd and Splash Damage Ltd are teaming up again to work | Read more »
BulkyPix Celebrates its 6th Anniversary...
BulkyPix Celebrates its 6th Anniversary with a Bunch of Free Games Posted by Jessica Fisher on December 19th, 2014 [ permalink ] BulkyPix has | Read more »
Indulge in Japanese cuisine in Cooking F...
Indulge in Japanese cuisine in Cooking Fever’s new sushi-themed update Posted by Simon Reed on December 19th, 2014 [ permalink ] Lithuanian developer Nordcurrent has yet again updated its restaurant simulat | Read more »
Badland Daydream Level Pack Arrives to C...
Badland Daydream Level Pack Arrives to Celebrate 20 Million Downloads Posted by Ellis Spice on December 19th, 2014 [ permalink ] | Read more »
Far Cry 4, Assassin’s Creed Unity, Desti...
Far Cry 4, Assassin’s Creed Unity, Destiny, and Beyond – AppSpy Takes a Look at AAA Companion Apps Posted by Rob Rich on December 19th, 2014 [ permalink ] These day | Read more »
A Bunch of Halfbrick Games Are Going Fre...
A Bunch of Halfbrick Games Are Going Free for the Holidays Posted by Ellis Spice on December 19th, 2014 [ permalink ] Universal App - Designed for iPhone and iPad | Read more »

Price Scanner via MacPrices.net

13-inch 2.6GHz Retina MacBook Pro on sale for...
Best Buy has lowered their price on the 2014 13″ 2.6GHz/128GB Retina MacBook Pro to $1149.99 on their online store for a limited time. That’s $150 off MSRP and the lowest price available for this... Read more
Kodak Returns to CES With New Consumer Produ...
Former photography colossus Kodak is returning to CES for the first time in three years where the Kodak booth (#21818 South Hall 1) will showcase a wide range of innovative, imaging-related products... Read more
Invaluable Launches New Eponymously -Named A...
Invaluable, the world’s largest online live auction marketplace, hhas announced the official launch of the Invaluable app for iPad, now available for download in the iTunes App Store. Invaluable... Read more
IDC Reveals Worldwide Mobile Enterprise Appli...
International Data Corporation (IDC) last week hosted the IDC FutureScape: Worldwide Mobile Enterprise Applications and Solutions 2015 Predictions Web conference. The session provided organizations... Read more
Hello Vino Wine App Launches “Safe Ride Home”...
Hello Vino has announced addition of a new “Get a Safe Ride Home” feature in its Food & Drink app with a direct connection to Uber, the technology platform that connects users with rides. The... Read more
DEVON-technologies Releases DEVONthink To Go...
Coeur d’Alene, Idaho based DEVON-technologies, LLC has updated DEVONthink To Go, its mobile companion to DEVONthink, to version 1.5. The update includes an iOS 8 extension, compatibility with the... Read more
The Apple Store offering free next-day shippi...
The Apple Store is now offering free next-day shipping on all in stock items if ordered before 12/23/14 at 10:00am PT. Local store pickup is also available within an hour of ordering for any in stock... Read more
It’s 1992 Again At Sony Pictures, Except For...
Techcrunch’s John Biggs interviewed a Sony Pictures Entertainment (SPE) employee, who quite understandably wished to remain anonymous, regarding post-hack conditions in SPE’s L.A office, explaining “... Read more
OtterBox Defender Series Case For iPad mini 3...
With their innovative Touch ID technology and ultrathin profile, the latest tranche of Apple iPads are more desirable than ever, and OtterBox has just announced the Defender Series custom-engineered... Read more
Holiday sales this weekend: MacBook Pros for...
 B&H Photo has new MacBook Pros on sale for up to $300 off MSRP as part of their Holiday pricing. Shipping is free, and B&H charges NY sales tax only: - 15″ 2.2GHz Retina MacBook Pro: $1699... Read more

Jobs Board

*Apple* Store Leader Program (US) - Apple, I...
…Summary Learn and grow as you explore the art of leadership at the Apple Store. You'll master our retail business inside and out through training, hands-on experience, Read more
Project Manager, *Apple* Financial Services...
**Job Summary** Apple Financial Services (AFS) offers consumers, businesses and educational institutions ways to finance Apple purchases. We work with national and Read more
*Apple* Retail - Multiple Positions (US) - A...
Sales Specialist - Retail Customer Service and Sales Transform Apple Store visitors into loyal Apple customers. When customers enter the store, you're also the Read more
*Apple* Retail - Multiple Positions (US) - A...
Sales Specialist - Retail Customer Service and Sales Transform Apple Store visitors into loyal Apple customers. When customers enter the store, you're also the Read more
*Apple* Retail - Multiple Positions (US) - A...
Job Description: Sales Specialist - Retail Customer Service and Sales Transform Apple Store visitors into loyal Apple customers. When customers enter the store, Read more
All contents are Copyright 1984-2011 by Xplain Corporation. All rights reserved. Theme designed by Icreon.