3 (This is not a most actively maintained file, but recommended
12 * Comments and Documentation
13 * Syntax-related Conventions
14 * Error and Warning Messages
21 To change the code at run-time the :SB-DEVEL feature is almost always
22 necessary, otherwise a lot of macros, functions, and variables do not
23 survive in the final image.
28 Preferred patch format is output from "git format-patch", including
31 The commit message should explain the why and how of the change. See
32 existing commit messages for examples -- for a truly trivial changes
33 little is needed, but in most cases more is better.
35 Please include test-cases in your patch if at all possible: if you're
36 not sure which file in tests/ to put your test case in, just pick one
37 that seems vaguely appropriate. See the "Writing Tests" section for
40 Please format your submission for ease of reading: unless the change
41 is whitespace only, avoid re-indenting code you are not touching, etc.
43 Unless your change is large and best understood as a series of
44 sequential changes, please send it in as single patch.
46 If your patch includes algorithmic changes, explain them. If your
47 patch uses a published algorithm, please include a link to the paper.
48 We aren't always as well-educated as we'd like to be...
50 Ready-to-apply patches should be submitted via Launchpad: please add
51 the tag "review" to the associated bug (create new bug with name if
52 there isn't one about the issue yet.)
54 Patches requiring more widespread discussion and feedback should be
55 sent to the sbcl-devel mailing list.
57 If you have any questions, feel free to ask them on sbcl-devel,
58 or the IRC channel #sbcl@irc.libera.chat.
63 See also PRINCIPLES and TLA files.
65 Most of the style hints in the Lisp FAQ apply.
67 When porting code we prefer code which factors dependencies into a set
68 of interface functions and constants and includes implementations of
69 the interface for the different systems.
71 Reader conditionals are frowed upon. Sometimes they're the least
72 of all evils, but think thrice.
74 grammatical fussiness:
75 Phrases are not capitalized.
76 Sentences are capitalized.
77 Periods terminate sentences.
78 Periods separate phrases from succeeding sentences, e.g.
79 ;;; the maximum number of transformations we'll make before
80 ;;; concluding we're in an infinite loop and bailing. This can
81 ;;; be changed, but it is an error to change it while we're
83 (defvar *max-n-transformations* 10)
84 Lisp in comments is capitalized.
85 Symbol names are capitalized.
88 Function documentation can be a description of what the function
90 ;;; Parse the arguments for a BDEFSTRUCT call, and return
91 ;;; (VALUES NAME DEFSTRUCT-ARGS MAKE-LOAD-FORM-FUN BDEFSTRUCT-STYPE),
92 ;;; where NAME is the name of the new type, DEFSTRUCT-ARGS is the
93 ;;; munged result suitable for passing on to DEFSTRUCT,
94 ;;; MAKE-LOAD-FORM-FUN is the make load form function, or NIL if
95 ;;; there's none, and BDEFSTRUCT-SUPERTYPE is the direct supertype
96 ;;; of the type if it is another BDEFSTRUCT-defined type, or NIL
98 (defun parse-bdefstruct-args (nameoid &rest rest)
100 or a remark about the function, e.g.
101 ;;; a helper function for BDEFSTRUCT in the #+XC-HOST case
102 (defun uncross-defstruct-args (defstruct-args)
104 If you're talking about what the function does, ordinarily you
105 should just say what the function does, e.g.
106 ;;; Return the first prime number greater than or equal to X.
107 (defun primify (x) ..)
108 instead of telling the reader that you're going to tell him what
109 the function does, e.g.
110 ;;; PRIMIFY returns the first prime number greater than or
112 (defun primify (x) ..)
114 ;;; When you call this function on X, you get back the first
115 ;;; prime number greater than or equal to X.
116 (defun primify (x) ..)
117 Documentation for public functions belongs in a docstring.
118 Documentation for internal functions belongs mostly in a comment.
120 In general, if you can express it in the code instead of the comments,
121 do so. E.g. the old CMUCL code has many comments above functions foo
124 If we were going to do something like that, we would prefer to do it by
127 (Instead, for various other reasons, we centralize all the exports
128 in package declarations.) The old "FOO -- interface" comments are bad
129 style because they duplicate information (and they illustrate one
130 of the evils of duplicating information by the way that they have
131 drifted out of sync with the code).
136 As mentioned in the "Patch Submissions" section, new features as well
137 as bug fixes should always be accompanied by smoke, unit or regression
138 tests as appropriate.
140 New tests should be added in the appropriate file in the "tests"
141 directory. Files named TOPIC.pure.lisp contain tests which do not have
142 side effects such as modifying the environment, defining functions or
143 variables or starting new threads while files named TOPIC.impure.lisp
144 contain test which do (impure test files are run each in a separate
147 All tests should use the WITH-TEST macro to associate a unique name
148 with each test (see below) and prevent a whole test file from going up
149 in smoke without an adequate description of the problem. The basic
152 (with-test (:name NAME)
155 where NAME is either a symbol or a list of symbols. All of these
156 symbols have to be in one of the packages KEYWORD, CL or SB-* to make
157 test names READable in SBCL images that did not read the test
158 files. For tests associated to Launchpad bugs (because a feature
159 requested in the LP bug has been implemented or a bug has been fixed),
160 the LP bug number should be a part of the test name:
162 (with-test (:name (coerce :invalid-target-type :bug-12345))
165 When a test can be associated to a name in the CL package or one of
166 the SB-* packages, that name should appear as a separate component in
167 the test name, like COERCE in the example above. This may be useful in
168 order to automatically associate functions and tests.
170 If possible, tests should not signal warnings during compilation or
171 runtime and should not produce output in order to not clutter the
172 output of the test runner and make real problems easier to identify.
177 There are a number of style practices on display in the code
178 which are not good examples to follow:
179 * using conditional compilation to support different architectures,
180 instead of factoring the dependencies into interfaces and providing
181 implementations of the interface for different architectures;
182 * in conditional compilation, using a common subexpression over and
183 over again, e.g. #+(OR X86 X86-64), when the important thing is
184 that the platform supports single-instruction CAS. If you have to
185 do something like that, define a new FOO feature, write #+FOO in
186 many places, and arrange for the FOO feature to be set once and
187 only once -- probably in make-config.sh. (That way future
188 maintainers won't curse you.)
189 * putting the defined symbol, and information about whether it's
190 exported or not, into the comments around the definition of the symbol;
191 * naming anything DO-FOO if it isn't an iteration macro
192 * not using a consistent abbreviation style in global names (e.g.
193 naming some things DEFINE-FOO and other things DEF-BAR, with
194 no rule to determine whether the abbreviation is used).
195 * using lots of single-colon package prefixes (distracting and hard
196 to read, and obstacles to reaching package nirvana where
197 package dependencies are a directed acyclic graph) or even
198 double-colon package prefixes (hard to understand and hard
199 to maintain). (One exception: I've sometimes been tempted to
200 add a CL: prefix to the definition of every CL symbol (e.g.
201 (DEFUN CL:CADDDR (..) ..) as reminders that they're required by
202 ANSI and can't be deleted no matter how obscure and useless some
203 of them might look.:-)
204 Many of these are common in the code inherited from CMUCL. We've
205 eliminated them in some places, but there's a *lot* of code inherited