1 Rules for the code in LyX
2 -------------------------
3 [updated from the C++STYLE distributed with the GNU C++ Standard]
5 The aim of this file is to serve as a guide for the developers, to aid us to
6 get clean and uniform code. This document is still incomplete.
8 We really like to have new developers joining the LyX Project. However
9 since we have had problems in the past with developers leaving the
10 project and their contributed code in a far from perfect state. Most
11 of this happened before that we really became aware of these issues,
12 but still, we don't want it to happen again. So we have put together
13 some guidelines and rules for the developers.
18 These guidelines should save us a lot of work while cleaning up the code and
19 help us to have quality code. LyX has been haunted by problems coming from
20 unfinished projects by people who have left the team. Those problems will
21 hopefully disappear if the code is easy to hand over to somebody else.
23 In general, if you want to contribute to the main source, we expect at least
26 - the most important rule first: kiss (keep it simple stupid), always
27 use a simple implementation in favor of a more complicated one.
28 This eases maintenance a lot.
29 - write good C++ code: Readable, well commented and taking advantage of the
30 OO model. Follow the formatting guidelines. See Formatting.
31 - adapt the code to the structures already existing in LyX, or in case that
32 you have better ideas, discuss them on the developer's list before writing
34 - take advantage of the C++ standard library. especially don't use
35 custom containers when a standard container is usable, learn to use
36 the algorithms and functors in the standard library.
37 - be aware of exceptions and write exception safe code. See Exceptions.
38 - document all variables, methods, functions, classes etc. We are
39 using the source documentation program doc++, a program that handles
40 javadoc syntax, to document sources. See Source Documentation.
41 - we have certain code constructs that we try to follow. See Code
48 It is implicitly understood that all patches contributed to The LyX
49 Project is under the Gnu General Public License, it you have a problem
50 with that, don't contribute code.
52 Also please don't just pup up out of the blue with a huge patch (or
53 small) that changes something substantial in LyX. Always discuss your
54 ideas with the developers on the developers mailing list.
56 When you create the patch, please use "diff -up" since we find that a
57 lot easier to read than the other diff formats. Also please do not
58 send patches that implements/fix several different things, several
59 patches is a much better option.
61 We also expect you to provide a ChangeLog entry with every patch, this
62 describes shortly what the patch is doing. The ChangeLog entry follows
65 1999-12-13 Lars Gullik Bjønnes <larsbj@lyx.org>
67 * src/support/lyxstring.C (find): assert bug fixed.
73 We have several guidelines on code constructs, some of these exists to
74 make the code faster, others to make the code clearer. Yet others
75 exists to make us able to take advantage of the strong type checking
78 - Declaration of variables should wait as long as possible. The rule
79 is: "Don't declare it until you need it." In C++ there are a lot of
80 user defined types, and these can very often be expensive to
81 initialize. This rule connects to the next rule too.
83 - Make the scope of a variable as small as possible.
85 - Prefer preincrement to postincrement whenever possible.
86 Preincrement has potential of being faster than postincrement. Just
87 thing about the obvious implementations of pre/post-increment. This
88 rule applies to decrement too.
96 - Try to minimize evaluation of the same code over and over. This is
97 aimed especially at loops.
99 Container::iterator end = large.end();
100 for (Container::iterator it = large.begin(); it != end; ++it) {
104 for (Container::iterator it = large.begin();
105 it != large.end(); ++it) {
109 - For functions and methods that return a non-POD type T, return T
110 const instead. This gives better type checking, and will give a
111 compiler warning when temporaries are used wrongly.
117 - Avoid using the default cases in switch statements unless you have
118 too. Use the correct type for the switch expression and let the
119 compiler ensure that all cases are exhausted.
127 case foo: ...; break;
128 case bar: ...; break;
129 default: ...; break; // not needed and would shadow a wrong use of Foo
135 Even if LyX currently is not using exceptions we need to be aware of
136 them. One important thing to realize is that you often do not have to
137 use throw,try or catch to be exception safe. Let's look at the
138 different types of exceptions safety: (These are taken from Herb
142 1. Basic guarantee: Even in the presence of exceptions thrown by T or
143 other exceptions, Stack objects don't leak resources.
144 Note that this also implies that the container will be
145 destructible and usable even if an exception is thrown wile
146 performing some container operation. However, if an exception
147 is thrown, the container will be in a consistent, but not
148 necessarily predictable, state. Containers that support the
149 basic guarantee can work safely in some settings.
151 2. Strong guarantee: If an operation terminates because of an
152 exception, program state will remain unchanged.
153 This always implies commit-or-rollback semantics, including
154 that no references or iterators into the container be
155 invalidated if an operation fails. For example, if a Stack
156 client calls Top and then attempts a Push that fails because
157 of an exception, then the state of the Stack object must be
158 unchanged and the reference returned from the prior call to
159 Top must still be valid. For more information on there
160 guarantees, see Dave Abrahams's documentation of the SGI
161 exception-safe standard library adaption at:
162 http://www.metabyte.com/~fbp/stl/eg_contract.html
164 Probably the most interesting point here is that when you
165 implement the basic guarantee, the strong guarantee often
166 comes for free. For example, in our Stack implementation,
167 almost everything we did was needed to satisfy just the basic
168 guarantee -- and what's presented above very nearly satisfies
169 the strong guarantee, with little of no extra work. Not half
170 bad, considering all the trouble we went to.
172 In addition to these two guarantees, there is one more
173 guarantee that certain functions must provide in order to make
174 overall exception safety possible:
176 3. Nothrow guarantee: The function will not emit an exception under any
178 Overall exception safety isn't possible unless certain
179 functions are guaranteed not to throw. In particular, we've
180 seen that this is true for destructors; later in this
181 miniseries, we'll see that it's also needed in certain helper
182 functions, such as Swap().
185 For all cases where we might be able to write exception safe functions
186 without using try, throw or catch we should do so. In particular we
187 should look over all destructors to ensure that they are as exception
190 Later when more compiler support exceptions sufficiently well we will
191 begin using them too. One reason for this is that the C++ standard
192 library actually requires exceptions, e.g. "new" will throw
193 bad_allocation if the requested memory is not available.
199 * Only one declaration on each line.
204 This is especially important when initialization is done at the same
209 string a("Lars"), b("Gullik"); // wrong
211 * Pointers and references
215 char *p = "flop"; // wrong
216 char &c = *p; // wrong
218 Some time ago we had a huge discussion on this subject and after
219 convincing argumentation from Asger this is what we decided. Also note
223 const char * p; // wrong
225 * Operator names and parentheses
228 operator == (type) // wrong
230 The == is part of the function name, separating it makes the
231 declaration look like an expression.
233 * Function names and parentheses
236 void mangle () // wrong
245 enum { one = 1, two = 2, three 3 }; // wrong
247 * Naming rules for classes
249 - Use descriptive but simple and short names. For stuff specific to LyX
250 use LyX as prefix. Some modules, like mathed or spellchecker, could have
252 [I am not so sure about the LyX prefix]
254 - Class names are usually capitalized, and function names lowercased.
255 Enums are named like Classes, enum values in CAPS.
257 - Long variables are named like thisLongVariableName.
259 New types are capitalized, so this goes for typedefs, classes, structs
264 - Please adapt the formatting of your code to the setting in LyX in that
265 particular file. Lars and Asger are slowly, but surely moving the source
266 towards Linux kernel style formatting, aka K&R style. We suggest that you
267 also do this, but this is NOT something that has been decided generally.
270 * Use existing structures
272 - Use string wherever possible. LyX will someday move to Unicode, and
273 that will be easy if everybody uses string now.
275 - Check out the filename and path tools in filetools.h
277 - Check out the string tools in lstring.h, and the SubString class
280 - Use the DebugStream class to report errors and messages using
281 the lyxerr instantiation.
283 [add description of other existing structures]
288 - Use this order for the access sections of your class: public,
289 protected, private. The public section is interesting for every
290 user of the class. The private section is only of interest for the
291 implementors of the class (you). [Obviously not true since this is
292 for developers, and we do not want one developer only to be able to
293 read and understand the implementation of class internals. Lgb]
295 - Avoid to declare global objects in the declaration file of the class.
296 If the same variable is used for all object, use a static member.
298 - Avoid global or static variables. An exception to this rule is
299 very private stuff like the math stack.
301 - Use the const keyword like this: char const * instead of const char *
302 because this is more logical.
307 - The documentation is generated from the header files.
308 - You document for the other developers, not for yourself.
309 - You should document what the function does, not the implementation.
310 - in the .C files you document the implementation.
311 - Single line description (///), multiple lines description (/** ... */)
312 - You make the documentation by doing "make srcdoc" in the root,
313 and then you'll find HTML in the srcdoc/ directory. Read with
314 Netscape for best results.
317 * NAMING RULES FOR USER-COMMANDS
319 Here's the set of rules to apply when a new command name is introduced:
321 1) Use the object.event order. That is, use `word-forward' instead of
323 2) Don't introduce an alias for an already named object. Same for events.
324 3) Forward movement or focus is called `forward' (not `right').
325 4) Backward movement or focus is called `backward' (not `left').
326 5) Upward movement of focus is called `up'.
327 6) Downward movement is called `down'.
328 7) The begin of an object is called `begin' (not `start').
329 8) The end of an object is called `end'.
332 * Using external GUI constructors (XForms fdesign)
334 - Fdesign generated files should not be changed at all. The only changes
335 needed are gettext, compability with 0.88 or when you have made your own
336 xforms objects and have just a dummy in the .fd file in place of your
337 own. In case you have to change the generated files for any of the
338 reasons above, you should provide a patch against the clean generated
339 file. Your callbacks must be in a separate file.
341 *************************************************************
343 How to create class interfaces.
344 (a.k.a How Non-Member Functions Improve Encapsulation)
345 ======================================================
347 I recently read an article by Scott Meyers in C/C++ Users
348 Journal (Vol.18,No.2), where he makes a strong case on how non-member
349 functions makes classes more encapsulated, not less. Just to skipping
350 to the core of this provides us with the following algorithm for
351 deciding what kind of function to add to a class interface:
353 - We need to add a function f to the class C's API.
355 if (f needs to be virtual)
356 make f a member function of C;
357 else if (f is operator>> or operator<<) {
358 make f a non-member function;
359 if (f needs access to non-public members of C)
360 make f a friend of C;
361 } else if (f needs type conversions on its left-most argument) {
362 make f a non-member function;
363 if (f needs access to non-public members of C)
364 make f a friend of C;
365 } else if (f can be implemented via C's public interface)
366 make f a non-member function;
368 make f a member function of C;
370 Unfortunately, to make the best use of this kind of Class API's we
371 need namespaces. As soon as Jean-Marc stops using gcc 2.8 and other
372 compilers seem more or less up to date on namespaces we will begin to
373 use them. _BUT_ we should begin to use the above algoritm ASAP. We
374 should also go through old code and apply this algorithm to the
375 existing member functions. That will help maintainability in the
378 (I'll fill in more from Scott Meyers article when time allows.)
383 [ExC++] Sutter, Herb. Exceptional C++: 47 engineering puzzles,
384 programming problems, and solutions. ISBN 0-201-61562-2