emacs.d/clones/lisp/www.cliki.net/Riastradh's Lisp Style Rules [for Common Lisp].html
2022-10-07 15:47:14 +02:00

588 lines
No EOL
36 KiB
HTML

<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>CLiki: Riastradh's Lisp Style Rules [for Common Lisp]</title>
<link rel="alternate" type="application/atom+xml" title="ATOM feed of edits to current article"
href="https://www.cliki.net/site/feed/article.atom?title=Riastradh%27s%20Lisp%20Style%20Rules%20%5Bfor%20Common%20Lisp%5D">
<link rel="stylesheet" href="static/css/style.css">
<link rel="stylesheet" href="static/css/colorize.css">
</head>
<body>
<span class="hidden">CLiki - Riastradh's Lisp Style Rules [for Common Lisp]</span>
<div id="content"><div id="content-area"><div id="article-title">Riastradh's Lisp Style Rules [for Common Lisp]</div><div id="article">Taylor R. Campbell wrote a style guide for Scheme code (<a href="http://mumble.net/~campbell/scheme/style.txt">link</a>). The <a href="Common&#32;Lisp.html" class="category">Common Lisp</a>-relevant sections are reposted below:<p><pre>Copyright (C) 2007 Taylor R. Campbell
All rights reserved.
Copies and quotations of this document are freely permitted, provided
that this copyright notice is preserved in copies and that the author
is cited in quotations.</pre><p>*** Spacing<p>If any text precedes an opening bracket or follows a closing bracket,
separate that text from that bracket with a space. Conversely, leave
no space after an opening bracket and before following text, or after
preceding text and before a closing bracket.<p><pre>
Unacceptable:
(foo(bar baz)quux)
(foo ( bar baz ) quux)
Acceptable:
(foo (bar baz) quux)
</pre>
Rationale: This is the same spacing found in standard typography of
European text. It is more aesthetically pleasing.<p>*** Line Separation<p>Absolutely do *not* place closing brackets on their own lines.
<pre>
Unacceptable:
(define (factorial x)
(if (&lt; x 2)
1
(* x (factorial (- x 1
)
)
)
)
)
Acceptable:
(define (factorial x)
(if (&lt; x 2)
1
(* x (factorial (- x 1)))))
</pre>
Rationale: The parentheses grow lonely if their closing brackets are
all kept separated and segregated.<p>**** Exceptions to the Above Rule Concerning Line Separation<p>Do not heed this section unless you know what you are doing. Its title
does *not* make the unacceptable example above acceptable.<p>When commenting out fragments of expressions with line comments, it may
be necessary to break a line before a sequence of closing brackets:
<pre>
(define (foo bar)
(list (frob bar)
(zork bar)
;; (zap bar)
))
</pre>
This is acceptable, but there are other alternatives. In Common Lisp,
one can use the read-time conditional syntax, `#+' or `#-', with a
feature conditional that is guaranteed to be false or true -- `#+(OR)'
or `#-(AND)' --; for example,
<pre>
(define (foo bar)
(list (frob bar)
(zork bar)
#+(or) (zap bar)))
</pre>
Finally, it is acceptable to break a line immediately after an opening
bracket and immediately before a closing bracket for very long lists,
especially in files under version control. This eases the maintenance
of the lists and clarifies version diffs. Example:
<pre>
(define colour-names ;Add more colour names to this list!
&#039;(
blue
cerulean
green
magenta
purple
red
scarlet
turquoise
))
</pre>
** Indentation and Alignment<p>The operator of any form, i.e. the first subform following the opening
round bracket, determines the rules for indenting or aligning the
remaining forms. Many names in this position indicate special
alignment or indentation rules; these are special operators, macros, or
procedures that have certain parameter structures.<p>If the first subform is a non-special name, however, then if the second
subform is on the same line, align the starting column of all following
subforms with that of the second subform. If the second subform is on
the following line, align its starting column with that of the first
subform, and do the same for all remaining subforms.<p>In general, Emacs will indent Lisp code correctly. Run `C-M-q'
(indent-sexp) on any code to ensure that it is indented correctly, and
configure Emacs so that any non-standard forms are indented
appropriately.
<pre>
Unacceptable:
(+ (sqrt -1)
(* x y)
(+ p q))
(+
(sqrt -1)
(* x y)
(+ p q))
Acceptable:
(+ (sqrt -1)
(* x y)
(+ p q))
(+
(sqrt -1)
(* x y)
(+ p q))
</pre>
Rationale: The columnar alignment allows the reader to follow the
operands of any operation straightforwardly, simply by scanning
downward or upward to match a common column. Indentation dictates
structure; confusing indentation is a burden on the reader who wishes
to derive structure without matching parentheses manually.<p>*** Non-Symbol Indentation and Alignment<p>The above rules are not exhaustive; some cases may arise with strange
data in operator positions.<p>**** Lists<p>Unfortunately, style varies here from person to person and from editor
to editor. Here are some examples of possible ways to indent lists
whose operators are lists:
<pre>
Questionable:
((car x) ;Requires hand indentation.
(cdr x)
foo)
((car x) (cdr x) ;GNU Emacs
foo)
Preferable:
((car x) ;Any Emacs
(cdr x)
foo)
((car x) (cdr x) ;Edwin
foo)
</pre>
Rationale: The operands should be aligned, as if it were any other
procedure call with a name in the operator position; anything other
than this is confusing because it gives some operands greater visual
distinction, allowing others to hide from the viewer's sight. For
example, the questionable indentation
<pre>
((car x) (cdr x)
foo)
</pre>
can make it hard to see that FOO and (CDR X) are both operands here
at the same level. However, GNU Emacs will generate that indentation
by default. (Edwin will not.)<p>**** Strings<p>If the form in question is meant to be simply a list of literal data,
all of the subforms should be aligned to the same column, irrespective
of the first subform.
<pre>
Unacceptable:
(&quot;foo&quot; &quot;bar&quot; &quot;baz&quot; &quot;quux&quot; &quot;zot&quot;
&quot;mumble&quot; &quot;frotz&quot; &quot;gargle&quot; &quot;mumph&quot;)
Questionable, but acceptable:
(3 1 4 1 5 9 2 6 5 3 5 8 9 7 9 3 2 3 8 4 6 2 6 4
3 3 8 3 2 7 9 5 0 2 8 8 4 1 9 7 1 6 9 3 9 9 3)
Acceptable:
(&quot;foo&quot; &quot;bar&quot; &quot;baz&quot; &quot;quux&quot; &quot;zot&quot;
&quot;mumble&quot; &quot;frotz&quot; &quot;gargle&quot; &quot;mumph&quot;)
(&quot;foo&quot;
&quot;bar&quot; &quot;baz&quot; &quot;quux&quot; &quot;zot&quot;
&quot;mumble&quot; &quot;frotz&quot; &quot;gargle&quot; &quot;mumph&quot;)
</pre>
Rationale: Seldom is the first subform distinguished for any reason,
if it is a literal; usually in this case it indicates pure data, not
code. Some editors and pretty-printers, however, will indent
unacceptably in the example given unless the second subform is on the
next line anyway, which is why the last way to write the fragment is
usually best.<p>** Names<p>Naming is subtle and elusive. Bizarrely, it is simultaneously
insignificant, because an object is independent of and unaffected by
the many names by which we refer to it, and also of supreme
importance, because it is what programming -- and, indeed, almost
everything that we humans deal with -- is all about. A full
discussion of the concept of name lies far outside the scope of this
document, and could surely fill not even a book but a library.<p>Symbolic names are written with English words separated by hyphens.
Scheme and Common Lisp both fold the case of names in programs;
consequently, camel case is frowned upon, and not merely because it is
ugly. Underscores are unacceptable separators except for names that
were derived directly from a foreign language without translation.
<pre>
Unacceptable:
XMLHttpRequest
foreach
append_map
Acceptable:
xml-http-request
for-each
append-map
</pre>
*** Funny Characters<p>There are several different conventions in different Lisps for the use
of non-alphanumeric characters in names.<p>**** Scheme<p>***** Question Marks: Predicates<p>Affix a question mark to the end of a name for a procedure whose
purpose is to ask a question of an object and to yield a boolean
answer. Such procedures are called `predicates'. Do not use a
question mark if the procedure may return any object other than a
boolean.<p> Examples: pair? procedure? proper-list?
Non-examples: member assoc any every<p>Pronounce the question mark as if it were the isolated letter `p'. For
example, to read the fragment (PAIR? OBJECT) aloud, say: `pair-pee
object.'<p>***** Exclamation Marks: Destructive Operations<p>Affix an exclamation mark to the end of a name for a procedure (or
macro) whose primary purpose is to modify an object. Such procedures
are called `destructive'.<p> Examples: set-car! append!<p>Avoid using the exclamation mark willy nilly for just *any* procedure
whose operation involves any kind of mutation or side effect; instead,
use the exclamation mark to identify procedures that exist *solely* for
the purpose of destructive update (e.g., SET-CAR!), or to distinguish a
destructive, or potentially destructive (in the case of linear-update
operations such as APPEND!), variant of a procedure of which there also
exists a purely functional variant (e.g., APPEND).<p>Pronounce the exclamation mark as `bang'. For example, to read the
fragment (APPEND! LIST TAIL) aloud, say: `append-bang list tail.'<p>***** Asterisks: Variants, Internal Routines, Mutable Globals<p>Affix an asterisk to the end of a name to make a variation on a theme
of the original name.<p> Example: let -&gt; let*<p>Prefer a meaningful name over an asterisk; the asterisk does not
explain what variation on the theme the name means.<p>Affix an asterisk to the beginning of a name to make an internal
routine for that name. Again, prefer a meaningful name over an
asterisk.<p>Affix asterisks to the beginning and end of a globally mutable
variable. This allows the reader of the program to recognize very
easily that it is badly written!<p>***** `WITH-' and `CALL-WITH-': Dynamic State and Cleanup<p>Prefix `WITH-' to any procedure that establishes dynamic state and
calls a nullary procedure, which should be the last (required)
argument. The dynamic state should be established for the extent of
the nullary procedure, and should be returned to its original state
after that procedure returns.<p> Examples: with-input-from-file with-output-to-file<p> Exception: Some systems provide a procedure (WITH-CONTINUATION
), which calls in the given
continuation, using that continuation's dynamic state. If
returns, it will return to , not to the continuation
of the call to WITH-CONTINUATION. This is acceptable, although the
name WITHIN-CONTINUATION (which one finds in MIT Scheme) is better.<p>Prefix `CALL-WITH-' to any procedure that calls a procedure, which
should be its last argument, with some arguments, and is either somehow
dependent upon the dynamic state or continuation of the program, or
will perform some action to clean up data after the procedure argument
returns. Generally, `CALL-WITH-' procedures should return the values
that the procedure argument returns, after performing the cleaning
action.<p> Examples:<p> - CALL-WITH-INPUT-FILE and CALL-WITH-OUTPUT-FILE both accept a
pathname and a procedure as an argument, open that pathname (for
input or output, respectively), and call the procedure with one
argument, a port corresponding with the file named by the given
pathname. After the procedure returns, CALL-WITH-INPUT-FILE and
CALL-WITH-OUTPUT-FILE close the file that they opened, and return
whatever the procedure returned.<p> - CALL-WITH-CURRENT-CONTINUATION is dependent on the continuation
with which it was called, and passes as an argument an escape
procedure corresponding with that continuation.<p> - CALL-WITH-OUTPUT-STRING, a common but non-standard procedure
definable in terms of OPEN-OUTPUT-STRING and GET-OUTPUT-STRING from
SRFI 6 (Basic String Ports), calls its procedure argument with an
output port, and returns a string of all of the output written to
that port. Note that it does not return what the procedure
argument returns, which is an exception to the above rule.<p>Generally, the distinction between these two classes of procedures is
that `CALL-WITH-...' procedures should not establish fresh dynamic
state and instead pass explicit arguments to their procedure arguments,
whereas `WITH-...' should do the opposite and establish dynamic state
while passing zero arguments to their procedure arguments.<p>** Comments<p>Write heading comments with at least four semicolons; see also the
section below titled `Outline Headings'.<p>Write top-level comments with three semicolons.<p>Write comments on a particular fragment of code before that fragment
and aligned with it, using two semicolons.<p>Write margin comments with one semicolon.<p>The only comments in which omission of a space between the semicolon
and the text is acceptable are margin comments.
<pre>
Examples:
;;;; Frob Grovel
;;; This section of code has some important implications:
;;; 1. Foo.
;;; 2. Bar.
;;; 3. Baz.
(define (fnord zarquon)
;; If zob, then veeblefitz.
(quux zot
mumble ;Zibblefrotz.
frotz))
</pre>
** General Layout<p>This section contains rules that the author has found generally helpful
in keeping his programs clean and presentable, though they are not
especially philosophically interesting.<p>Contained in the rationale for some of the following rules are
references to historical limitations of terminals and printers, which
are now considered aging cruft of no further relevance to today's
computers. Such references are made only to explain specific measures
chosen for some of the rules, such as a limit of eighty columns per
line, or sixty-six lines per page. There is a real reason for each of
the rules, and this real reason is not intrinsically related to the
historical measures, which are mentioned only for the sake of
providing some arbitrary measure for the limit.<p>*** File Length<p>If a file exceeds five hundred twelve lines, begin to consider
splitting it into multiple files. Do not write program files that
exceed one thousand twenty-four lines. Write a concise but
descriptive title at the top of each file, and include no content in
the file that is unrelated to its title.<p> Rationale: Files that are any larger should generally be factored
into smaller parts. (One thousand twenty-four is a nicer number than
one thousand.) Identifying the purpose of the file helps to break it
into parts if necessary and to ensure that nothing unrelated is
included accidentally.<p>*** Top-Level Form Length<p>Do not write top-level forms that exceed twenty-one lines, except for
top-level forms that serve only the purpose of listing large sets of
data. If a procedure exceeds this length, split it apart and give
names to its parts. Avoid names formed simply by appending a number
to the original procedure's name; give meaningful names to the parts.<p> Rationale: Top-level forms, especially procedure definitions, that
exceed this length usually combine too many concepts under one name.
Readers of the code are likely to more easily understand the code if
it is composed of separately named parts. Simply appending a number
to the original procedure's name can help only the letter of the
rule, not the spirit, however, even if the procedure was taken from a
standard algorithm description. Using comments to mark the code with
its corresponding place in the algorithm's description is acceptable,
but the algorithm should be split up in meaningful fragments anyway.<p> Rationale for the number twenty-one: Twenty-one lines, at a maximum
of eighty columns per line, fits in a GNU Emacs instance running in a
24x80 terminal. Although the terminal may have twenty-four lines,
three of the lines are occupied by GNU Emacs: one for the menu bar
(which the author of this guide never uses, but which occupies a line
nevertheless in a vanilla GNU Emacs installation), one for the mode
line, and one for the minibuffer's window. The writer of some code
may not be limited to such a terminal, but the author of this style
guide often finds it helpful to have at least four such terminals or
Emacs windows open simultaneously, spread across a twelve-inch laptop
screen, to view multiple code fragments.<p>*** Line Length<p>Do not write lines that exceed eighty columns, or if possible
seventy-two.<p> Rationale: Following multiple lines that span more columns is
difficult for humans, who must remember the line of focus and scan
right to left from the end of the previous line to the beginning of
the next line; the more columns there are, the harder this is to do.
Sticking to a fixed limit helps to improve readability.<p> Rationale for the numbers eighty and seventy-two: It is true that we
have very wide screens these days, and we are no longer limited to
eighty-column terminals; however, we ought to exploit our wide
screens not by writing long lines, but by viewing multiple fragments
of code in parallel, something that the author of this guide does
very often. Seventy-two columns leave room for several nested layers
of quotation in email messages before the code reaches eighty
columns. Also, a fixed column limit yields nicer printed output,
especially in conjunction with pagination; see the section
`Pagination' below.<p>*** Blank Lines<p>Separate each adjacent top-level form with a single blank line (i.e.
two line breaks). If two blank lines seem more appropriate, break the
page instead. Do not place blank lines in the middle of a procedure
body, except to separate internal definitions; if there is a blank
line for any other reason, split the top-level form up into multiple
ones.<p> Rationale: More than one blank line is distracting and sloppy. If
the two concepts that are separated by multiple blank lines are
really so distinct that such a wide separator is warranted, then
they are probably better placed on separate pages anyway; see the
next section, `Pagination'.<p>*** Pagination<p>Separate each file into pages of no more than sixty-six lines and no
fewer than forty lines with form feeds (ASCII #x0C, or ^L, written in
Emacs with `C-q C-l'), on either side of which is a single line break
(but not a blank line).<p> Rationale: Keeping distinct concepts laid out on separate pages
helps to keep them straight. This is helpful not only for the
writer of the code, but also for the reader. It also allows readers
of the code to print it onto paper without fiddling with printer
settings to permit pages of more than sixty-six lines (which is the
default number for many printers), and pagination also makes the
code easier to navigate in Emacs, with the `C-x [' and `C-x ]' keys
(`backward-page' and `forward-page', respectively). To avoid
excessively small increments of page-by-page navigation, and to
avoid wasting paper, each page should generally exceed forty lines.<p> `C-x l' in Emacs will report the number of lines in the page on which
the point lies; this is useful for finding where pagination is
necessary.<p>*** Outline Headings<p>Use Emacs's Outline Mode to give titles to the pages, and if
appropriate a hierarchical structure. Set `outline-regexp' (or
`outline-pattern' in Edwin) to "\f\n;;;;+ ", so that each form feed
followed by an line break followed by at least four semicolons and a
space indicates an outline heading to Emacs. Use four semicolons for
the highest level of headings in the hierarchy, and one more for each
successively nested level of hierarchy.<p> Rationale: Not only does this clarify the organization of the code,
but readers of the code can then navigate the code's structure with
Outline Mode commands such as `C-c C-f', `C-c C-b', `C-c C-u', and
`C-c C-d' (forward, backward, up, down, respectively, headings).<p>*** Dependencies<p>When writing a file or module, minimize its dependencies. If there are
too many dependencies, consider breaking the module up into several
parts, and writing another module that is the sum of the parts and that
depends only on the parts, not their dependencies.<p> Rationale: A fragment of a program with fewer dependencies is less
of a burden on the reader's cognition. The reader can more easily
understand the fragment in isolation; humans are very good at local
analyses, and terrible at global ones.<p>** Naming<p>This section requires an elaborate philosophical discussion which the
author is too ill to have the energy to write at this moment.<p>Compose concise but meaningful names. Do not cheat by abbreviating
words or using contractions.<p> Rationale: Abbreviating words in names does not make them shorter;
it only makes them occupy less screen space. The reader still must
understand the whole long name. This does not mean, however, that
names should necessarily be long; they should be descriptive. Some
long names are more descriptive than some short names, but there are
also descriptive names that are not long and long names that are not
descriptive. Here is an example of a long name that is not
descriptive, from SchMUSE, a multi-user simulation environment
written in MIT Scheme:<p> frisk-descriptor-recursive-subexpr-descender-for-frisk-descr-env<p> Not only is it long (sixty-four characters) and completely
impenetrable, but halfway through its author decided to abbreviate
some words as well!<p>Do not write single-letter variable names. Give local variables
meaningful names composed from complete English words.<p> Rationale: It is tempting to reason that local variables are
invisible to other code, so it is OK to be messy with their names.
This is faulty reasoning: although the next person to come along and
use a library may not care about anything but the top-level
definitions that it exports, this is not the only audience of the
code. Someone will also want to read the code later on, and if it is
full of impenetrably terse variable names without meaning, that
someone will have a hard time reading the code.<p>Give names to intermediate values where their expressions do not
adequately describe them.<p> Rationale: An `expression' is a term that expresses some value.
Although a machine needs no higher meaning for this value, and
although it should be written to be sufficiently clear for a human to
understand what it means, the expression might mean something more
than just what it says where it is used. Consequently, it is helpful
for humans to see names given to expressions.<p> Example: A hash table HASH-TABLE maps foos to bars; (HASH-TABLE/GET
HASH-TABLE FOO #F) expresses the datum that HASH-TABLE maps FOO to,
but that expression gives the reader no hint of any information
concerning that datum. (LET ((BAR (HASH-TABLE/GET FOO #F))) ...)
gives a helpful name for the reader to understand the code without
having to find the definition of HASH-TABLE.<p> Index variables such as i and j, or variables such as A and D naming
the car and cdr of a pair, are acceptable only if they are completely
unambiguous in the scope. For example,<p><pre> (do ((i 0 (+ i 1)))
((= i (vector-length vector)))
(frobnicate (vector-ref vector i)))</pre><p> is acceptable because the scope of i is very clearly limited to a
single vector. However, if more vectors are involved, using more
index variables such as j and k will obscure the program further.<p>Avoid functional combinators, or, worse, the point-free (or
`point-less') style of code that is popular in the Haskell world. At
most, use function composition only where the composition of functions
is the crux of the idea being expressed, rather than simply a procedure
that happens to be a composition of two others.<p> Rationale: Tempting as it may be to recognize patterns that can be
structured as combinations of functional combinators -- say, `compose
this procedure with the projection of the second argument of that
other one', or (COMPOSE FOO (PROJECT 2 BAR)) --, the reader of the
code must subsequently examine the elaborate structure that has been
built up to obscure the underlying purpose. The previous fragment
could have been written (LAMBDA (A B) (FOO (BAR B))), which is in
fact shorter, and which tells the reader directly what argument is
being passed on to what, and what argument is being ignored, without
forcing the reader to search for the definitions of FOO and BAR or
the call site of the final composition. The explicit fragment
contains substantially more information when intermediate values are
named, which is very helpful for understanding it and especially for
modifying it later on.<p> The screen space that can be potentially saved by using functional
combinators is made up for by the cognitive effort on the part of the
reader. The reader should not be asked to search globally for usage
sites in order to understand a local fragment. Only if the structure
of the composition really is central to the point of the narrative
should it be written as such. For example, in a symbolic integrator
or differentiator, composition is an important concept, but in most
code the structure of the composition is completely irrelevant to the
real point of the code.<p>If a parameter is ignored, give it a meaningful name nevertheless and
say that it is ignored; do not simply call it `ignored'.<p>In Common Lisp, variables can be ignored with (DECLARE (IGNORE ...)).
Some Scheme systems have similar declarations, but the portable way to
ignore variables is just to write them in a command context, where
their values will be discarded, preferably with a comment indicating
this purpose:<p><pre> (define (foo x y z)
x z ;ignore
(frobnitz y))</pre><p> Rationale: As with using functional combinators to hide names,
avoiding meaningful names for ignored parameters only obscures the
purpose of the program. It is helpful for a reader to understand
what parameters a procedure is independent of, or if someone wishes
to change the procedure later on, it is helpful to know what other
parameters are available. If the ignored parameters were named
meaninglessly, then these people would be forced to search for call
sites of the procedure in order to get a rough idea of what
parameters might be passed here.<p>When naming top-level bindings, assume namespace partitions unless in a
context where they are certain to be absent. Do not write explicit
namespace prefixes, such as FOO:BAR for an operation BAR in a module
FOO, unless the names will be used in a context known not to have any
kind of namespace partitions.<p> Rationale: Explicit namespace prefixes are ugly, and lengthen names
without adding much semantic content. Common Lisp has its package
system to separate the namespaces of symbols; most Schemes have
mechanisms to do so as well, even if the RnRS do not specify any. It
is better to write clear names which can be disambiguated if
necessary, rather than to write names that assume some kind of
disambiguation to be necessary to begin with. Furthermore, explicit
namespace prefixes are inadequate to cover name clashes anyway:
someone else might choose the same namespace prefix. Relegating this
issue to a module system removes it from the content of the program,
where it is uninteresting.<p>** Comments<p>Write comments only where the code is incapable of explaining itself.
Prefer self-explanatory code over explanatory comments. Avoid
`literate programming' like the plague.<p> Rationale: If the code is often incapable of explaining itself, then
perhaps it should be written in a more expressive language. This may
mean using a different programming language altogether, or, since we
are talking about Lisp, it may mean simply building a combinator
language or a macro language for the purpose. `Literate programming'
is the logical conclusion of languages incapable of explaining
themselves; it is a direct concession of the inexpressiveness of the
computer language implementing the program, to the extent that the
only way a human can understand the program is by having it rewritten
in a human language.<p>Do not write interface documentation in the comments for the
implementation of the interface. Explain the interface at the top of
the file if it is a single-file library, or put that documentation in
another file altogether. (See the `Documentation' section below if the
interface documentation comments grow too large for a file.)<p> Rationale: A reader who is interested only in the interface really
should not need to read through the implementation to pick out its
interface; by putting the interface documentation at the top, not
only is such a reader's task of identifying the interface made
easier, but the implementation code can be more liberally commented
without fear of distracting this reader. To a reader who is
interested in the implementation as well, the interface is still
useful in order to understand what concepts the implementation is
implementing.<p> Example: <p> In this example of a single-file library implementing the skip list
data structure, the first page explains the purpose and dependencies
of the file (which are useful for anyone who intends to use it, even
though dependencies are really implementation details), and the next
few pages explain the usage of skip lists as implemented in that
file. On the first page of implementation, `Skip List Structure',
there are some comments of interest only to a reader who wishes to
understand the implementation; the same goes for the rest of the
file, none of which must a reader read whose interest is only in the
usage of the library.<p>Avoid block comments (i.e. #| ... |#). Use S-expression comments (`#;'
in Scheme, with the expression to comment on the next line; `#+(OR)' or
`#-(AND)' in Common Lisp) to comment out whole expressions. Use blocks
of line comments for text.<p> Rationale: Editor support for block comments is weak, because it
requires keeping a detailed intermediate parse state of the whole
buffer, which most Emacsen do not do. At the very least, #|| ... ||#
is better, because most Emacsen will see vertical bars as symbol
delimiters, and lose trying to read a very, very long symbol, if they
try to parse #| ... |#, whereas they will just see two empty symbols
and otherwise innocuous text between them if they try to parse #||
... ||#. In any case, in Emacs, `M-x comment-region RET', or `M-;'
(comment-dwim), is trivial to type.<p> The only standard comments in Scheme are line comments. There are
SRFIs for block comments and S-expression comments, but support for
them varies from system to system. Expression comments are not hard
for editors to deal with because it is safe not to deal with them at
all; however, in Scheme S-expression comments, which are written by
prefixing an expression with `#;', the expression to be commented
should be placed on the next line. This is because editors that do
not deal with them at all may see the semicolon as the start of a
line comment, which will throw them off. Expression comments in
Common Lisp, however, are always safe.<p> In Common Lisp, the two read-time conditionals that are guaranteed to
ignore any form following them are `#+(OR)' and `#-(AND)'. `#+NIL'
is sometimes used in their stead, but, while it may appear to be an
obviously false conditional, it actually is not. The feature
expressions are read in the KEYWORD package, so NIL is read not as
CL:NIL, i.e. the boolean false value, but as :NIL, a keyword symbol
whose name happens to be `NIL'. Not only is it not read as the
boolean false value, but it has historically been used to indicate a
feature that might be enabled -- in JonL White's New Implementation
of Lisp! However, the New Implementation of Lisp is rather old these
days, and unlikely to matter much...until Alastair Bridgewater writes
Nyef's Implementation of Lisp.<p>** Documentation<p>On-line references and documentation/manuals are both useful for
independent purposes, but there is a very fine distinction between
them. Do not generate documentation or manuals automatically from the
text of on-line references.<p> Rationale: /On-line references/ are quick blurbs associated with
objects in a running Lisp image, such as documentation strings in
Common Lisp or Emacs Lisp. These assume that the reader is familiar
with the gist of the surrounding context, but unclear on details;
on-line references specify the details of individual objects.<p> /Documentation/ and /manuals/ are fuller, organized, and cohesive
documents that explain the surrounding context to readers who are
unfamiliar with it. A reader should be able to pick a manual up and
begin reading it at some definite point, perusing it linearly to
acquire an understanding of the subject. Although manuals may be
dominated by reference sections, they should still have sections that
are linearly readable to acquaint the reader with context.<p><hr>
<a href="Document.html" class="category">Document</a></div></div>
<div id="footer" class="buttonbar"><ul><li><a href="Riastradh's&#32;Lisp&#32;Style&#32;Rules&#32;[for&#32;Common&#32;Lisp].html">Current version</a></li>
<li><a href="https://www.cliki.net/site/history?article=Riastradh%27s%20Lisp%20Style%20Rules%20%5Bfor%20Common%20Lisp%5D">History</a></li>
<li><a href="https://www.cliki.net/site/backlinks?article=Riastradh%27s%20Lisp%20Style%20Rules%20%5Bfor%20Common%20Lisp%5D">Backlinks</a></li><li><a href="https://www.cliki.net/site/edit-article?title=Riastradh%27s%20Lisp%20Style%20Rules%20%5Bfor%20Common%20Lisp%5D&amp;from-revision=3775823103">Edit</a></li><li><a href="https://www.cliki.net/site/edit-article?create=t">Create</a></li></ul></div>
</div>
<div id="header-buttons" class="buttonbar">
<ul>
<li><a href="https://www.cliki.net/">Home</a></li>
<li><a href="https://www.cliki.net/site/recent-changes">Recent Changes</a></li>
<li><a href="CLiki.html">About</a></li>
<li><a href="Text&#32;Formatting.html">Text Formatting</a></li>
<li><a href="https://www.cliki.net/site/tools">Tools</a></li>
</ul>
<div id="search">
<form action="https://www.cliki.net/site/search">
<label for="search_query" class="hidden">Search CLiki</label>
<input type="text" name="query" id="search_query" value="" />
<input type="submit" value="search" />
</form>
</div>
</div>
<div id="pageheader">
<div id="header">
<span id="logo">CLiki</span>
<span id="slogan">the common lisp wiki</span>
<div id="login"><form method="post" action="https://www.cliki.net/site/login">
<label for="login_name" class="hidden">Account name</label>
<input type="text" name="name" id="login_name" class="login_input" />
<label for= "login_password" class="hidden">Password</label>
<input type="password" name="password" id="login_password" class="login_input" />
<input type="submit" name="login" value="login" id="login_submit" /><br />
<div id="register"><a href="https://www.cliki.net/site/register">register</a></div>
<input type="submit" name="reset-pw" value="reset password" id="reset_pw" />
</form>
</div>
</div>
</div>
</body></html>