Commit | Line | Data |
---|---|---|
cbed66ad | 1 | Notes on coding style |
1528431b | 2 | |
cbed66ad MW |
3 | * General |
4 | ||
5 | ** Layout | |
6 | ||
7 | Lines are 77 characters at most, except for strange special effects. | |
8 | Don't ask. This is not negotiable, though. Don't try to tell me that | |
9 | your monitor is very wide so you can read longer lines. My monitor is | |
10 | likely at least as wide. On the other hand, most lines are easily short | |
11 | enough to fit in my narrow columns, so the right hand side of a wide | |
12 | window would be mostly blank. This seems wasteful to me, when I could | |
13 | fill that space with more code. | |
14 | ||
15 | Horizontal whitespace for layout purposes -- i.e., indentation and | |
16 | alignment, rather than just separating words -- consists of as many tabs | |
17 | as possible, followed by as many spaces as necessary to reach the target | |
18 | column. Tab stops occur at every eight columns. You can tell this | |
19 | because when you cat a file to your terminal, that's how the tabs | |
20 | appear. Editors which disagree about this are simply wrong. | |
21 | ||
22 | My indentation quantum is usually two columns. It seems that some | |
23 | modern editors are deeply confused, and think that tab width and | |
24 | indentation quantum are the same thing, but they aren't. Such broken | |
25 | editors will make a hopeless mess of my code. If you have the | |
26 | misfortune to use such an editor, maybe you could contribute patches to | |
27 | fix it. | |
28 | ||
29 | ||
30 | * Lisp style | |
31 | ||
32 | ** Language subset and extensions | |
1528431b MW |
33 | |
34 | None of ANSI Common Lisp is off-limits. | |
35 | ||
828cb3b1 MW |
36 | I think my Lisp style is rather more imperative in flavour than most |
37 | modern Lisp programmers. It's probably closer to historical Lisp | |
a1b51e19 MW |
38 | practice in that regard, even though I wasn't writing Lisp back then. A |
39 | lot of this is because I don't assume that the Lisp implementation | |
40 | handles tail calls properly: Common Lisp is not Scheme. | |
828cb3b1 | 41 | |
1528431b MW |
42 | I make extensive use of CLOS, and macros. On a couple of occasions I've |
43 | made macros which use CLOS generic function dispatch to compute their | |
44 | expansions. The parser language is probably the best example of this in | |
828cb3b1 MW |
45 | the codebase. |
46 | ||
a1b51e19 MW |
47 | I like hairy ~format~ strings. I've intentionally opted to leave them |
48 | as challenges to the reader rather than explain them. | |
1528431b MW |
49 | |
50 | I've avoided hairy ~loop~ for the most part, not because I dislike it | |
51 | strongly but because others do and I don't find that it wins big enough | |
52 | for the fight to be worthwhile. | |
53 | ||
54 | I only use ~&aux~ lambda-list parameters in ~defstruct~ BOA | |
55 | constructors, for special effects. | |
56 | ||
57 | I use ~car~, not ~first~, and ~cdr~, not ~rest~. Similarly, I use | |
58 | ~cadr~, not ~second~, and I'm not afraid to use ~cddr~ or ~cadar~. | |
59 | ||
60 | Similarly, I've not used ~elt~, preferring to know what kind of sequence | |
61 | I'm dealing with, or using the built-in sequence functions. | |
62 | ||
63 | I'm happy to use ~1+~, and I like the brevity of ~1-~ enough to use it | |
64 | despite its terrible name. | |
65 | ||
66 | There are no reader syntax extensions in the code. This is because I | |
67 | couldn't think of any way they'd be especially helpful, and not because | |
68 | I'm in any way opposed to them. | |
69 | ||
70 | The main translator, in the ~SOD~ package, tries to assume very little | |
71 | beyond ANSI Common Lisp and what's included in just about every serious | |
72 | implementation: specifically, MOP introspection, and Gray streams. | |
73 | There's intentionally no MOP intercession. | |
74 | ||
7e55d099 MW |
75 | The frontend additionally makes use of ~cl-launch~, but the dependency |
76 | is actually quite weak, and it could be replaced with a different, maybe | |
1528431b MW |
77 | implementation-specific, mechanism fairly easily. I'm keen to take |
78 | patches which improve frontend portability. | |
79 | ||
80 | I'm more tolerant of extensions and external dependencies in the test | |
81 | suite, which makes additional use of ~xlunit~. Running the test suite | |
82 | isn't essential to getting the translator built, so this isn't as much | |
83 | of a problem. | |
84 | ||
cbed66ad | 85 | ** Layout |
1528431b MW |
86 | |
87 | I pretty much let Emacs indent my code for me, based on information | |
88 | collected by SLIME. Some exceptions: | |
89 | ||
90 | + DSLs (e.g., the parser language) have their own space of macros | |
91 | which Emacs doesn't understand and for the most part I haven't | |
92 | bothered to teach it. | |
93 | ||
94 | + Emacs sometimes does a bad job with hairy ~loop~ and requires manual | |
95 | fixing. Since I don't use hairy ~loop~ much, this isn't a major | |
96 | problem. | |
97 | ||
f458e64e MW |
98 | + Emacs indents lambda lists really badly. I often prefer to put the |
99 | entire lambda list on its own line than to split it. If I have to | |
100 | split a simple lambda list, without lambda-list keywords, I just | |
101 | align the start of each subsequent line with the start of the first | |
102 | argument. I break hairy lambda lists before lambda-list keywords, | |
103 | and the start of a subsequent line aligns with the first argument | |
104 | name following the lambda-list keyword which begins the group, so | |
105 | that the lambda-list keyword stands out. | |
106 | ||
107 | : (defun many-arguments (first second third | |
108 | : fourth fifth) | |
109 | : ...) | |
110 | ||
111 | : (defun hairy-arguments (first second third | |
112 | : &optional fourth fifth | |
113 | : sixth | |
114 | : &rest others) | |
115 | : ...) | |
116 | ||
117 | I don't know what I'd do if I had a hairy lambda list with so many | |
118 | mandatory positional arguments that I had to split them. So far, | |
119 | this situation hasn't come up. | |
120 | ||
1528431b MW |
121 | Lisp code does have a tendency to march across to the right quite |
122 | rapidly given a chance. I have a number of strategies for dealing with | |
123 | this. | |
124 | ||
125 | + Break a long nested calculation into pieces, giving names to the | |
126 | intermediate results, in a ~let*~ form. | |
127 | ||
7e55d099 MW |
128 | + Hoist deeply nested complex computations out into ~flet~ or |
129 | ~labels~, and then invoke them from inside whatever complicated | |
1528431b MW |
130 | conditional mess was needed to decide what to do. |
131 | ||
132 | + Shrug my shoulders and let code dribble down the right hand side for | |
133 | a bit. | |
134 | ||
cbed66ad | 135 | ** Packages and exporting |
1528431b MW |
136 | |
137 | A package collects symbols which are given meanings in one or more | |
138 | source files. If a package's code is all in one file, then the package | |
139 | definition can be put in that file too; otherwise I put it in its own | |
140 | file. | |
141 | ||
142 | I don't put ~:export~ in package definitions. Instead, I scatter calls | |
143 | to the ~export~ function throughout the code, right next to where the | |
144 | relevant symbol is defined. This has three important advantages. | |
145 | ||
146 | + You can tell, when you're reading the code which defines ~foo~, | |
147 | whether ~foo~ is exported and therefore a defined part of the | |
148 | package interface. | |
149 | ||
150 | + When you know that you're writing a thing which will form part of | |
151 | the package interface, you don't have to go off and edit some other | |
152 | file to export it. | |
153 | ||
154 | + A master list of exported symbols becomes a merge hazard: if two | |
155 | different branches add symbols to nearby pieces of the master list | |
156 | then you get a merge conflict for no especially good reason. | |
157 | ||
158 | There's an apparent disadvantage: there's no immediately visible master | |
159 | list of exported symbols. But that's not a big problem: | |
160 | ||
161 | : (loop for s being the external-symbols of pkg collect s) | |
162 | ||
163 | See ~doc/list-symbols.lisp~ for more sophisticated reporting. (In | |
164 | particular, this identifies what kind of thing(s) each external symbol | |
165 | names.) | |
166 | ||
cbed66ad | 167 | ** Comments and file structuring |
1528431b MW |
168 | |
169 | A file starts with a big ~;;;~ comment bearing the Emacs ~-*-lisp-*-~ | |
170 | marker, a quick description, and copyright and licensing boilerplate. I | |
171 | don't use four-semicolon comments, and I only use ~#|~ ... ~|#~ for | |
172 | special effects. | |
173 | ||
174 | Then there's package stuff. There may be a ~cl:defpackage~ form (with | |
175 | explicit package qualifier) if the relevant package doesn't have its own | |
a1b51e19 MW |
176 | package definition file. I use gensyms to name packages: strings don't |
177 | seem right, and symbols would leak into some unrelated package. | |
1528431b MW |
178 | |
179 | Then there's ~cl:in-package~. Like ~defpackage~, I use a gensym to name | |
180 | the package. I can't think offhand of a good reason to have a file with | |
7e55d099 MW |
181 | sections `in' more than one package. So, the ~in-package~ form goes at |
182 | the top of the file, before the first section header. If sections are | |
183 | going to end up in separate packages, I think I'd put a ~cl:in-package~ | |
184 | at the top of each section in case I wanted to reorder them. | |
1528431b MW |
185 | |
186 | The rest of the file consists of Lisp code. I don't use page boundaries | |
187 | ~^L~ to split files up. Instead, I use big banner comments for this: | |
188 | ||
189 | : ;;;-------------------------------------------------------------------------- | |
190 | : ;;; Section title. | |
191 | ||
192 | Sections don't usually have internal comments, but if they did they'd | |
193 | also be ~;;;~ comments. | |
194 | ||
195 | Almost all definitions get documentation strings. I've tried to be | |
196 | consistent about formatting. | |
197 | ||
198 | + Docstring lines are 77 characters or less. | |
199 | ||
200 | + The first line gives a summary of what the thing does. The summary, | |
201 | together with the SLIME-generated synopsis, is likely enough to | |
202 | remind you what the thing does. | |
203 | ||
204 | + The rest of the lines are indented by three spaces, and explain | |
205 | carefully what the thing does and what all the parameters mean. | |
206 | ||
207 | Smallish functions and macros don't usually need any further | |
208 | commentary. Big functions often need to be split into bitesize pieces | |
209 | with their own internal ~;;~ comments. The idea is that these comments | |
210 | should explain the code's overall strategy to the reader, and help them | |
211 | figure out how a piece fits into that strategy. | |
212 | ||
213 | Winged, single ~;~ comments are very rare. | |
214 | ||
215 | Files end, as a result of long tradition, with a comment | |
216 | ||
217 | : ;;;----- That's all, folks -------------------------------------------------- | |
218 | ||
cbed66ad | 219 | ** Macro style |
1528431b MW |
220 | |
221 | I don't mind complicated macros if they're doing something worthwhile. | |
222 | They need to have good documentation strings, though. | |
223 | ||
224 | That said, where possible I've tried to factor macros into an actual | |
225 | macro providing the syntactic sugar, and a function which receives the | |
226 | parameters and $\eta$-expanded forms, and does the actual work. | |
227 | ||
228 | It's extremely bad taste for a macro to evaluate its evaluable | |
229 | parameters in any order other than strictly left to right, or to | |
230 | evaluate them more than once. | |
231 | ||
cbed66ad | 232 | ** Data structures |
1528431b MW |
233 | |
234 | I've tended to be happy with plain lists for homogeneous-ish | |
235 | collections. Strongly heterogeneous collections (other than input | |
236 | syntax, destructured using ~defmacro~ or ~destructuring-bind~) I've | |
237 | tended to make a proper data type for. | |
238 | ||
239 | My first instinct when defining a new structure is to use ~defclass~. | |
240 | While it's annoyingly verbose, it has the immense benefit over | |
241 | ~defstruct~ that it's safe to redefine CLOS classes in a running image | |
242 | without the world breaking, and I usually find it necessary to add or | |
243 | change slots while I'm working on new code. Once a piece of code has | |
244 | settled down and I have a good feel for what my structure is actually | |
245 | doing, I might switch the ~defclass~ for a ~defstruct~. Several | |
246 | questions influence my decision. | |
247 | ||
248 | + Do slot accesses need to be really fast? My usual Lisp | |
249 | implementations aggressively optimize ~defstruct~ accessor | |
250 | functions. | |
251 | ||
7e55d099 | 252 | + Have I subclassed my class? While I can move over a |
1528431b MW |
253 | single-inheritance tree using ~:include~, it seems wrong to do this |
254 | most of the time. Also, I'd be precluding subclasses from multiple | |
255 | inheritance, and I'd either have to prohibit subclassing by | |
256 | extensions or have to commit to ~defstruct~ in the documentation. | |
257 | In general, I'm much happier committing to ~defclass~. | |
258 | ||
259 | + Are there methods specialized on my class? Again, structure classes | |
260 | make fine method specializers, but it doesn't seem right. | |
261 | ||
262 | Apart from being hard to redefine, ~defstruct~ does a pretty good job of | |
263 | making a new structure type. I tend to tidy up a few rough edges. | |
264 | ||
265 | + The default predicate always has ~-p~ appended. If the class name | |
266 | is a single word, then I'll explicitly name the predicate with a | |
267 | simple ~p~ suffix. For example, ~ship~ would have the predicate | |
a51bf71a | 268 | ~shipp~, rather than ~ship-p~. |
1528431b MW |
269 | |
270 | + If there are slots I can't default then I'll usually provide a BOA | |
271 | constructor which sets them from required parameters; other slots | |
272 | I'll set from optional or keyword parameters according to my taste | |
273 | and judgement. | |
274 | ||
275 | + Slots mustn't be given names which are external in any package. | |
276 | Unfortunately, slot names are used in constructing accessor names, | |
277 | and sometimes the right accessor name involves a prohibited symbol. | |
278 | I've mostly addressed this by naming the slot ~%foo~, and then | |
279 | providing inline reader and writer functions. (CLOS class | |
280 | definitions don't have this problem because you get to set the | |
281 | accessor function names independently of the slot names.) | |
282 | ||
283 | + BOA constructors are strange. You can set the initial slots based | |
284 | on an arbitrary computation on the provided parameters, but you have | |
285 | to roll up your sleeves and mess with ~&aux~ parameters to pull it | |
286 | off. | |
287 | ||
cbed66ad | 288 | ** Naming |
1528431b MW |
289 | |
290 | I'm a traditionalist in some ways, and one of the reasons I like Lisp is | |
291 | the richness of its history and tradition. | |
292 | ||
293 | In other languages, I tend to use single- or two-letter names for | |
294 | variables and structure slots; not so much in Lisp. Other languages | |
295 | express more using punctuation, so the names stand out easily; I find | |
296 | that short names can be lost more easily in Lisp. | |
297 | ||
298 | I've also tended to go for fairly prosaic names, taking my inspiration | |
299 | from the CLOS MOP. While I mourn the loss of whimsical names like | |
300 | ~haulong~ and ~haipart~, I've tried to avoid inventing more of them. | |
301 | ||
f458e64e MW |
302 | There's a convention, which I think comes from ML, of using ~_~ where a |
303 | binding occurrence of a variable name is expected, to signify that that | |
304 | the corresponding value is to be discarded. Common Lisp, alas, doesn't | |
305 | have such a convention. Instead, there's a sequence of silly names used | |
306 | with the same intention, and the bindings are then explicitly ignored | |
307 | with a declaration. The names begin ~hunoz~, ~hukairz~, and (I think) | |
308 | ~huaskt~. | |
1528431b | 309 | |
cbed66ad | 310 | ** Declarations |
1528431b MW |
311 | |
312 | The code is light on declarations, other than ~ignore~ and similar used | |
313 | to muffle warnings. The macros try to do sensible things with | |
314 | declarations, and I think they succeed fairly well, but there might be | |
315 | bugs and rough edges. I know that some are just broken because, for | |
316 | actual correctness, declarations provided by the caller need to be split | |
317 | up into a number of different parts of the expansion, which in turn | |
318 | requires figuring out what the declarations mean and which bindings | |
319 | they're referring to. That's not completely impossible, assuming that | |
f458e64e | 320 | there aren't implementation-specific declarations with crazy syntax |
1528431b MW |
321 | mixed in there, but it's more work than seems worthwhile. |
322 | ||
323 | ||
324 | * COMMENT Emacs cruft | |
325 | ||
326 | #+LATEX_CLASS: strayman | |
327 | ||
328 | ## LocalWords: CLOS ish destructure destructured accessor specializers | |
a1b51e19 | 329 | ## LocalWords: accessors DSLs gensym gensyms |
1528431b MW |
330 | |
331 | ## Local variables: | |
332 | ## mode: org | |
333 | ## End: |