Liberty Eiffel Programming Environment FAQ

Frequently Asked Questions

General purpose questions about the SmartEiffel project

Is there a SmartEiffel mailing list?

Yes! SmartEiffel users and developers can share their experiences and ideas by subscribing to the SmartEiffel official mailing list.

Please, note that this list is not moderated. Please, stay correct on that list.

Is it possible to have the complete Eiffel source code of SmartEiffel?

Since it is the GNU Eiffel Compiler, the complete source code of SmartEiffel is of course in the distribution. The source code for additional libraries is also provided. See also this FAQ question which is related to licensing.

Is it possible to use SmartEiffel for large/commercial applications?

It is indeed possible to use SmartEiffel for a large application. An Eiffel compiler is a really big project and SmartEiffel itself is fully written in Eiffel. SmartEiffel is completely free and any private company can use SmartEiffel freely, and distribute (or sell) freely the products made with it. They do not have to pay royalties.

Also note that only classes which are closely related to the compiler itself are under GPL (actually, only the classes of directory SmartEiffel/tools are under GPL). The other classes are not under GPL in order to let people completely free (i.e. all classes of SmartEiffel/lib).

Always keep in mind that SmartEiffel doesn't come with any warranty (please read the COPYING file in the distribution). As explained in the header of non-GPL files, the only important thing is to kept the header unaltered when the corresponding source file is provided in your product. For extended support, donations as well as consulting, see here.

What is SmallEiffel compared to SmartEiffel?

SmallEiffel is the former name of the SmartEiffel project. We changed because we thought the compiler had become smart enough ;)

For a list of changes between the last version of SmallEiffel and the first one of SmartEiffel, have a look there.

How can I help?

The best way to help SmartEiffel and its users is probably to pick up some area you're interested in, and develop with SmartEiffel your own library, application, tool, extension, etc., releasing it to other users.

To avoid duplicate efforts, we suggest you have a look to what other people are already doing around SmartEiffel and help them if something that's of interest to you is already under work. There are actually very good things in there, some we were not even aware of, when they were under development ! :)

You may have a look to the future plans and have a look into the bugs database (access via the bug report page.

Finally, you may also consider funding the development of particular features you would like to have.

Another very good way to help us is to follow the bug report guidelines when you find some problem with SmartEiffel.

Why don't you change this and add that?! It would be much better/cooler/whatever!

People must understand that we can't always do everything. We simply can't. Because we don't have the time. Whether we like it or not, we also have other things to do than provide free stuff, modify our compiler and/or libraries to please everybody. We do as much as we can, but we don't do miracles, sorry.

Since SmartEiffel is free of charge and open-source, people who do need things we don't provide can always implement them and have them benefit everybody. A good way to do this is to start working on it, and ask other people (i.e. not the SmartEiffel team ;) ) to join and help. See the How can I help? question.

Alternatively, someone or some company who does need us to implement something may always consider funding a bit the development of SmartEiffel... After all, we've even heard that some people were selling software and making a bit of money with it... ;)))

How do I keep informed about SmartEiffel?

The best way is to keep an eye on our web pages, more especially on the What's New? page and the History of Changes.

People who want detailed information may also decide to stay tuned to the SmartEiffel mailing-list or even look at the bug-tracking system.

What documentation is provided with SmartEiffel?

The documentation provided with SmartEiffel is a transcript of what you can find on the original SmartEiffel Web site, at http://SmartEiffel.loria.fr. It is related only to the use and internals of SmartEiffel (yes, we know we still have to improve it ;) ).

For information and documentation about the Eiffel language, check the links on our Internet resources page. Note that we are not aware of any complete Eiffel language manual freely available on the Web (yet?).

Why don't you post more messages in newsgroups and/or mailing lists?

First, because we strongly believe that too much information kills information. Scientists call this "cognitive overload". :)

Second, because we don't have the time. It takes an awful amount of time to follow discussions, whatever their quality. We try to do that. But it's even more time-consuming to be part of them. So, we often have to choose between posting/mailing, and working directly on SmartEiffel. Since our mailboxes tend to overflood, we generally choose the latter :)

Is it difficult to switch from some commercial Eiffel compiler to SmartEiffel?

If your original Eiffel software only uses simple types like INTEGER, STRING, ARRAY, BOOLEAN, CHARACTER and DOUBLE, it is usually very simple to modify your code in order to use SmartEiffel.

It is a little bit difficult for simple input/output (used with predefined io) because some features have different names. If your original software heavily relies for example on the EiffelBase library, it may be very difficult. For example, one must keep in mind that SmartEiffel.ARRAY inherit SmartEiffel.COLLECTION and that ISE library also have a class COLLECTION. By the way, subclasses of ISE.COLLECTION cannot be used. The ISE.LINKED_LIST can be used in conjunction with SmartEiffel.ARRAY because ISE.LINKED_LIST do not inherit ISE.COLLECTION (no clash).

Questions about Languages and/or Object-Oriented Programming

What is the difference between the Static and the Dynamic Type?

The static type of a variable is determined at compile-time, the dynamic type during run-time. The dynamic type must conform to the static type (it may and often will be the same as the static type).

E.g.: A variable might be declared to refer to an instance of class FRUIT (a_fruit: FRUIT), so the static type is FRUIT, but might be assigned an object of class APPLE (a_fruit := an_apple), which becomes the dynamic type. Obviously, APPLE has to be a descendant of FRUIT in this example.

What is type prediction?

Type prediction is when the compiler attempts to predict an expression's dynamic type at compile time.

Questions about the Eiffel language of the SmartEiffel project

Why is SmartEiffel case-sensitive?

In fact, like most computer languages, Eiffel does distinguish between upper and lower case. Our decision is simply guided by the goals of legibility, organisation and simplification.

As far as keywords are concerned, it is easier to distinguish them when they are always written in the same way. Thus, we always write loop, rather than Loop or even LoOp. We have to get into good habits from now on in other matters too: this rule is going to be generalised so as to be even stricter in future. From now on, lets all get into the habit of writing Current, Void or True, for example.

Upper case is also used to distinguish class names from other names. Thus, we can know that FOO is a class name and foo is not, without having to look elsewhere in the source code.

A final advantage of the upper/lower case distinction is that it allows more precise error messages. What makes work simpler for humans also simplifies it for compilers.

What is the semantics (meaning) of a manifest string?

It can certainly be important to know what really happens when you define a constant STRING attribute by writing, for example:

Message: STRING is "abc"

that constant attribute definition is in fact a compact way of writing:

Message: STRING is
   once
      create Result.make(3)
      Result.extend('a')
      Result.extend('b')
      Result.extend('c')
   end

You realise, of course, that that function definition is a once function and consequently one string and one string only will be created, whatever happens. When you are investigating performance, you also have to know the difference between:

string := "abc"

and

string := once "abc"

You have probably guessed that, if the instruction without once is in a loop, a new string is created each time that the line in question is executed!

The once expression form above is only valid for manifest STRINGs and UNICODE_STRINGs. Here is an example of a manifest UNICODE_STRING (note the capital U):

unicode_string := once U"abc"

The [compile_to_c] command's -manifest_string_trace option allows you to locate unwanted non-once manifest strings.

Can you explain again the difference between conformance and covariance?

Our paper " Conformance of agents in the Eiffel language has a note p.137 that looks surprising.

Note that the previous rules define conformance rules, this has nothing to do with covariance or contravariance.

The concepts are distinct, even though there is a relation between them.

Covariance is defined in terms of conformance. But assignment is also defined in terms of conformance.

That's exactly, and only, what our paper defines: assignment rules. Those rules were not defined anywhere before. It's important: it's not that we disagreed with any earlier position. We just filled a hole in the specification.

Also note that, even if the agent types notation uses square brackets, it's totally unrelated to the generic classes type notation; therefore generic classes rules cannot apply (except if proved otherwise). An agent type is not a generic type, anymore than a tuple type is a generic type. Those are distinct concepts, only with a similar notation.

Conformance is fundamental in any typed language; on the other hand covariance is "just" an Eiffel extra. Important, sure, but not as fundamental as conformance. Our paper never speaks of covariance, except in the note quoted above, and explained further here. Maybe the paper should have explained the subtlety in so many words.

What is a CATCALL?

In the Eiffel world CATCALL is about type safety problems which are still present in the language. In Eiffel, CATCALL is a short-hand for Changing Availability or Type of CALLs.

Availability is about the exportation status of some method. For example, when a method is redefined, it's exportation status can be changed too.

Type of CALLs is about some possibly covariant redefinition.

What is SCOOP ?

In Eiffel, SCOOP refers to language support for distributed programming. In the Eiffel world, it is an acronym for Simple Concurrent Object-Oriented Programming. SCOOP is not yet included in SmartEiffel and is certainly its most important missing feature.

Questions about libraries which comes along with SmartEiffel

How should I read a file?

In order to read a file, read a character and then use last_character ONLY if if end_of_input has not been reached while reading the character.
Before each read (except the very first), you have to test end_of_input, because all read_* procedures require not end_of_input. But why? This require means that if some previous read failed because end of input has been reached, then it is not valid to try to read again. Example:

file: TEXT_FILE_READ; file_name: STRING
...
-- Assumming here that `file' and `file_name' are not Void:
file.connect_to(file_name)
if file.is_connected then
   from
      file.read_character
      until
         file.end_of_input
      loop
         io.put_character(file.last_character)
         file.read_character
      end
      file.disconnect
end

Questions about the tools of SmartEiffel

Is it possible to do incremental compilation with SmartEiffel?

Because of the SmartEiffel type inference mechanism, SmartEiffel always produces all needed C files from scratch. As old C files are automatically saved, only modified C files are recompiled. See man/compile for details.

Is there a mechanism to pre-compile libraries?

No, there is no such mechanism in SmartEiffel. But if you're concerned about compilation speed, don't worry, pre-computed libraries are not the only way to be fast ! Just try SmartEiffel, and you'll see :)

Is it possible to use the Boehm-Demers-Weiser garbage collector with SmartEiffel?

Yes.
You just have to disable the generation of the GC by SmartEiffel (option -no_gc) and replace it by the BDW library. Note that the BDW GC can be easily added because SmartEiffel only uses malloc/realloc/calloc.
You can download the BDW library from : 

http://www.hpl.hp.com/personal/Hans_Boehm/gc/

How to integrate the Boehm-Demers-Weiser garbage collector with SmartEiffel (from a message of Dietmar Wolz) : 

1. Make the gc.a file from the BDW package. Then move this gc.a file to /usr/lib/libgc.a.  
2. Move gc.h the to working directory and add the following lines: 

#define malloc(n) GC_malloc(n) 
#define calloc(m,n) GC_malloc((m)*(n)) 
#define realloc(p,n) GC_realloc((p),(n)) 
#define free(p) GC_free(p) 
3. instead of calling 

    "compile -o executable -boost -O root" 
now call  
    "compile -o executable -boost -O root -no_gc -lgc -include gc.h" 

How to customise the Garbage Collector for a new architecture?

If your architecture needs special handling to get the GC working (because the stack is not contiguous, because some registers are not written on the stack on `setjmp'...) then you need to implement function `mark_stack_and_registers' for your system in the file SmartEiffel/sys/runtime/c/gc_lib.c.

If you get some message telling you that the stack direction is wrong, then you should change macro definition to use the other generic code (there is one generic code for each stack order).

In order to check the GC, you should be able to run all files of the SmartEiffel/misc/benchmarks/gc directory.

How is SmartEiffel compiled?

With Eiffel optimisation options -boost and -no_gc. The garbage collector is indeed useless on the SmartEiffel commands: since SmartEiffel did not include a GC in its first versions, we were very careful about memory when we developed it.

With C compilation optimisations turned on (it depends on the C compiler used; we generally use gcc). The resulting executables are stripped.

Questions about the SmartEiffelWiki

Why do you have a Table of Contents page in your Wiki?

The reason for that is that we want to be able to print a paper version of all the information available in the wiki. The Table Of Contents page indicates the order of pages to be printed. Note that the new command to produce the great book is not yet written.

Why are all pages name in English?

The question may look strange for this English wiki. But note that we also have a French version.

As you may have noticed, all pages names are in English, even in the French wiki. For example, while reading the French Table des matières, the name of the page is still Table of Contents. Well, it is just to simplify the wiki itself. Note that it is straightforward to change for example an English link to the corresponding French link and conversely.

How to add figures created with tgif?

At the moment, all the figures in the Big Book are created by the program tgif, because it lets you produce .png files for the on-line version (the wiki) as well as .eps files for producing LaTeX documents (for paper versions).

Since figures are required to be editable by everyone, the source files for the tgif figures, the .obj files, are put on the server. It is the rule that for every file Foo.png on the server there must also be the source tgif file whose name must be Foo.obj.

To put a file on the server, use the Upload file button on the left-hand menu, which brings up the page of the same name, the upload file page.

To go the opposite way, when you want to download the tgif source file for a figure, you have to click on the Special pages button in the left-hand menu, in order to get to the page which contains the Image list link. Then click on the description of the .obj file that you want so as to be able to do the usual Save Link Target on the .obj file.

Is it essential to use tgif for figures?

For the moment, all the figures have been created with tgif, mainly for historical reasons, but also because this program works well, runs under Linux and can transform its schemas both into .png format and also, which is absolutely essential, into .eps format. Furthermore, tgif's source format, the .obj format, is an easily legible and portable text format.

We do not have anything against any other program, so long as that program is also able to produce the .png format for the wiki as well as a format that is compatible with LaTeX. The format of the source files must also be legible and portable. Finally, the program in question must also run under Linux and be free software.

As you will have realised, we really do want to keep life simple and so we prefer to restrict the number of programs used to produce figures. If you really want us to accept another program, contact us by using the most appropriate page of our Wiki. (which is what?)