SCM

SCM Repository

[inlinedocs] Annotation of /pkg/inlinedocs/R/parsers.R
ViewVC logotype

Annotation of /pkg/inlinedocs/R/parsers.R

Parent Directory Parent Directory | Revision Log Revision Log


Revision 248 - (view) (download)

1 : tdhock 110 ### combine lists or character strings
2 :     combine <- function(x,y)UseMethod("combine")
3 :    
4 :     ### combine character strings by pasting them together
5 :     combine.character <- function(x,y)
6 :     paste(x,y,sep="\n")
7 :    
8 :     ### combine lists by adding elements or adding to existing elements
9 :     combine.list <- function(x,y){
10 : tdhock 200 toadd <- if(".overwrite"%in%names(y)){
11 :     y <- y[names(y)!=".overwrite"]
12 :     rep(TRUE,length(y))
13 :     }else{
14 :     !names(y)%in%names(x)
15 :     }
16 : tdhock 185 toup <- names(y)[!toadd]
17 : tdhock 110 x[names(y)[toadd]] <- y[toadd]
18 :     for(up in toup)x[[up]] <- combine(x[[up]],y[[up]])
19 : tdhock 200 x
20 : tdhock 110 ### A list, same type as x, but with added elements from y.
21 :     }
22 :    
23 :     ### Prefix for code comments used with grep and gsub.
24 :     prefix <- "^[ \t]*###[ \t]"
25 :    
26 :     decomment <- function
27 :     ### Remove comment prefix and join lines of code to form a
28 :     ### documentation string.
29 :     (comments
30 :     ### Character vector of prefixed comment lines.
31 :     ){
32 : tdhock 185 gsub(prefix,"",comments)
33 : tdhock 110 ### String without prefixes or newlines.
34 :     }
35 :    
36 : tdhock 162 forall <- function
37 : tdhock 158 ### For each object in the package that satisfies the criterion
38 :     ### checked by subfun, parse source using FUN and return the resulting
39 : tdhock 87 ### documentation list.
40 :     (FUN,
41 :     ### Function to apply to each element in the package.
42 :     subfun=function(x)TRUE
43 :     ### Function to select subsets of elements of the package, such as
44 :     ### is.function. subfun(x)==TRUE means FUN will be applied to x and
45 :     ### the result will be returned.
46 :     ){
47 : tdhock 138 FUN <- FUN
48 :     f <- function(objs,docs,...){
49 : tdhock 87 objs <- objs[sapply(objs,subfun)]
50 :     L <- list()
51 : tdhock 162 on.exit(cat(sprintf("Parser Function failed on %s\n",N)))
52 : tdhock 204 for(N in union(names(docs),names(objs))){
53 : tdhock 87 o <- objs[[N]]
54 : tdhock 89 L[[N]] <- FUN(src=attr(o,"source"),
55 :     name=N,objs=objs,o=o,docs=docs,doc=docs[[N]],...)
56 : tdhock 87 }
57 : tdhock 162 on.exit()## remove warning message
58 : tdhock 87 L
59 :     }
60 : tdhock 138 class(f) <- c("allfun","function")
61 :     f
62 : tdhock 87 ### A Parser Function.
63 :     }
64 :    
65 : tdhock 138 ### Print method for functions constructed using forall.
66 :     print.allfun <- function(x,...){
67 :     e <- environment(x)
68 :     cat("Function to apply to every element.\nselector:")
69 :     print(e$subfun)
70 :     cat("processor:")
71 :     print(e$FUN)
72 :     }
73 :    
74 : tdhock 87 ### For each function in the package, do something.
75 :     forfun <- function(FUN)forall(FUN,is.function)
76 :    
77 : tdhock 163 kill.prefix.whitespace <- function
78 :     ### Figure out what the whitespace preceding the example code is, and
79 :     ### then delete that from every line.
80 :     (ex
81 :     ### character vector of example code lines.
82 :     ){
83 :     tlines <- gsub("\\s*","",ex)
84 :     ##tlines <- gsub("#.*","",tlines)
85 :     prefixes <- unique(gsub("\\S.*","",ex[tlines!=""]))
86 :     FIND <- prefixes[which.min(nchar(prefixes))]
87 :     ## Eliminate leading tabulations or 2/4 spaces
88 :     sub(FIND, "", ex)
89 :     ### Character vector of code lines with preceding whitespace removed.
90 :     }
91 :    
92 : tdhock 138 examples.after.return <- function
93 :     ### Get examples from inline definitions after return()
94 :     ### PhG: this does not work well! Think of these situations:
95 :     ### 1) You have multiple return() in the code of your function,
96 :     ### 2) You have return() appearing is some example code, ...
97 :     ### I can hardly propose a hack here. The whole code of the function
98 :     ### must be parsed, and one must determine which one is the last line
99 :     ### of code that is actually executed.
100 :     ###
101 :     ### I make two propositions here
102 :     ### 1) to keep the same mechanism that has the advantage of simplicity
103 :     ### but to use a special tag
104 :     ### ##examples<< or #{{{examples to separate
105 :     ### function code from examples explicitly, and
106 :     ### 2) to place the example in an "ex" attribute
107 :     ### attached to the function
108 :     ### (see next parser). That solution will be also interesting for
109 :     ### documenting datasets, something not done yet by inlinedocs!
110 :     (src,name="",...) {
111 :     ## Look for the examples mark
112 :     m <- grep("^\\s*(##examples<<|#\\{\\{\\{examples)", src)
113 :     if (!length(m)) return(list())
114 :     if (length(m) > 1)
115 :     warning("More than one examples tag for ", name,
116 :     ". Taking the last one")
117 :     m <- m[length(m)]
118 :     ## Look for the lines containing return value comments just before
119 :     r <- grep("\\s*### ", src[1:(m-1)])
120 :     if (!length(r)) {
121 :     value <- NULL
122 :     } else {
123 :     ## Only take consecutive lines before the mark
124 :     keep <- rev((m - rev(r)) == 1:length(r))
125 :     if (!any(keep)) {
126 :     value <- NULL
127 :     } else {
128 :     value <- decomment(src[r[keep]])
129 :     }
130 :     }
131 :     ## Collect now the example code beneath the mark
132 :     ex <- src[(m + 1):(length(src) - 1)]
133 :     ## Possibly eliminate a #}}} tag
134 :     ex <- ex[!grepl("#}}}", ex)]
135 :     ## Eliminate leading tabulations or four spaces
136 : tdhock 163 ex <- kill.prefix.whitespace(ex)
137 : tdhock 138 ## Add an empty line before and after example
138 :     ex <- c("", ex, "")
139 :     ## Return examples and value
140 : tdhock 185 list(examples = ex, value = value)
141 : tdhock 138 }
142 :    
143 : tdhock 175 prefixed.lines <- structure(function(src,...){
144 : tdhock 138 ### The primary mechanism of inline documentation is via consecutive
145 :     ### groups of lines matching the specified prefix regular expression
146 :     ### "\code{^### }" (i.e. lines beginning with "\code{### }") are
147 :     ### collected as follows into documentation sections:\describe{
148 :     ### \item{description}{group starting at line 2 in the code}
149 :     ### \item{arguments}{group following each function argument}
150 :     ### \item{value}{group ending at the penultimate line of the code}}
151 :     ### These may be added to by use of the \code{##<<} constructs
152 :     ### described below.
153 :     clines <- grep(prefix,src)
154 :     if(length(clines)==0)return(list())
155 :     bounds <- which(diff(clines)!=1)
156 :     starts <- c(1,bounds+1)
157 :     ends <- c(bounds,length(clines))
158 :     ## detect body of function using paren matching
159 : tdhock 159 code <- gsub("#.*","",src)
160 :     f <- function(ch)cumsum(nchar(gsub(sprintf("[^%s]",ch),"",code)))
161 : tdhock 138 parens <- f("(")-f(")")
162 :     body.begin <- which(diff(parens)<0 & parens[-1]==0)+2
163 : tdhock 175 if(length(body.begin)==0)body.begin <- 1 ## rare cases
164 : tdhock 138 is.arg <- function(){
165 : tdhock 159 gres <- grep("^\\s*#",src[start-1],perl=TRUE)
166 :     0 == length(gres) && start<=body.begin
167 :     }
168 : tdhock 138 res <- list()
169 :     for(i in seq_along(starts)){
170 :     start <- clines[starts[i]]
171 :     end <- clines[ends[i]]
172 : tdhock 175 lab <- if(all(grepl("^\\s*#",src[end:(length(src)-1)])))"value"
173 : tdhock 138 else if(start==2)"description"
174 :     else if(is.arg()){
175 :     ##twutz: strip leading white spaces and brackets and ,
176 :     arg <- gsub("^[ \t(,]*", "", src[start - 1])
177 :     arg <- gsub("^([^=,]*)[=,].*", "\\1", arg)
178 :     ##twutz: remove trailing whitespaces
179 :     arg <- gsub("^([^ \t]*)([ \t]+)$","\\1",arg)
180 :     arg <- gsub("...", "\\dots", arg, fix = TRUE)
181 :     paste("item{",arg,"}",sep="")
182 :     } else {
183 :     next;
184 :     }
185 :     res[[lab]] <- decomment(src[start:end])
186 :     }
187 :     res
188 : tdhock 175 },ex=function(){
189 :     test <- function
190 : tdhock 191 ### the description
191 : tdhock 175 (x,
192 :     ### the first argument
193 :     y ##<< another argument
194 :     ){
195 :     5
196 :     ### the return value
197 :     ##seealso<< foobar
198 : tdhock 138 }
199 : tdhock 175 src <- attr(test,"source")
200 :     prefixed.lines(src)
201 :     extract.xxx.chunks(src)
202 :     })
203 : tdhock 138
204 : tdhock 154 extract.xxx.chunks <- function # Extract documentation from a function
205 : tdhock 87 ### Given source code of a function, return a list describing inline
206 :     ### documentation in that source code.
207 : tdhock 154 (src,
208 :     ### The source lines of the function to examine, as a character
209 :     ### vector.
210 :     name.fun="(unnamed function)",
211 : tdhock 87 ### The name of the function/chunk to use in warning messages.
212 : tdhock 154 ...
213 :     ### ignored.
214 : tdhock 95 ){
215 : tdhock 87 res <- list()
216 :     ##details<< For simple functions/arguments, the argument may also be
217 :     ## documented by appending \code{##<<} comments on the same line as the
218 :     ## argument name. Mixing this mechanism with \code{###} comment lines for
219 :     ## the same argument is likely to lead to confusion, as the \code{###}
220 :     ## lines are processed first.
221 :     #arg.pat <- paste("^[^=,#]*?([\\w\\.]+)\\s*([=,].*|\\)\\s*)?",
222 :     # "<<\\s*(\\S.*?)\\s*$",
223 :     # sep="##") # paste avoids embedded trigger fooling the system
224 :     #tw: removed first comma
225 :     arg.pat <- paste("^[^=#]*?([\\w\\.]+)\\s*([=,].*|\\)\\s*)?",
226 :     "<<\\s*(\\S.*?)\\s*$",
227 :     sep="##") # paste avoids embedded trigger fooling the system
228 :    
229 :     skeleton.fields <- c("alias","details","keyword","references","author",
230 :     "note","seealso","value","title","description",
231 :     "describe","end")
232 :     ##details<< Additionally, consecutive sections of \code{##} comment
233 :     ## lines beginning with \code{##}\emph{xxx}\code{<<} (where
234 :     ## \emph{xxx} is one of the fields: \code{alias}, \code{details},
235 :     ## \code{keyword}, \code{references}, \code{author}, \code{note},
236 :     ## \code{seealso}, \code{value}, \code{title} or \code{description})
237 :     ## are accumulated and inserted in the relevant part of the .Rd
238 :     ## file.
239 :     ##
240 :     ## For \code{value}, \code{title}, \code{description} and function
241 :     ## arguments, these \emph{append} to any text from "prefix"
242 :     ## (\code{^### }) comment lines, irrespective of the order in the
243 :     ## source.
244 :     ##
245 :     ## When documenting S4 classes, documentation from \code{details}
246 :     ## sections will appear under a section \code{Objects from the Class}. That
247 :     ## section typically includes information about construction methods
248 :     ## as well as other description of class objects (but note that the
249 :     ## class Slots are documented in a separate section).
250 :    
251 :     ## but this should not appear, because separated by a blank line
252 :     extra.regexp <- paste("^\\s*##(",paste(skeleton.fields,collapse="|"),
253 :     ")<<\\s*(.*)$",sep="")
254 :     cont.re <- "^\\s*##\\s*"
255 :     in.describe <- 0
256 :     first.describe <- FALSE
257 :     k <- 1
258 :     in.chunk <- FALSE
259 :     end.chunk <- function(field,payload)
260 :     {
261 :     if ( "alias" == field ){
262 :     ##note<< \code{alias} extras are automatically split at new lines.
263 :     payload <- gsub("\\n+","\\}\n\\\\alias\\{",payload,perl=TRUE)
264 :     chunk.sep <- "}\n\\alias{"
265 :     } else if ( "keyword" == field ){
266 :     ##keyword<< documentation utilities
267 :     ##note<< \code{keyword} extras are automatically split at white space,
268 :     ## as all the valid keywords are single words.
269 :     payload <- gsub("\\s+","\\}\n\\\\keyword\\{",payload,perl=TRUE)
270 :     chunk.sep <- "}\n\\keyword{"
271 :     } else if ( "title" == field ){
272 :     chunk.sep <- " "
273 :     } else if ( "description" == field ){
274 :     chunk.sep <- "\n"
275 :     } else {
276 :     ##details<< Each separate extra section appears as a new
277 :     ## paragraph except that: \itemize{\item empty sections (no
278 :     ## matter how many lines) are ignored;\item \code{alias} and
279 :     ## \code{keyword} sections have special rules;\item
280 :     ## \code{description} should be brief, so all such sections
281 :     ## are concatenated as one paragraph;\item \code{title} should
282 :     ## be one line, so any extra \code{title} sections are
283 :     ## concatenated as a single line with spaces separating the
284 :     ## sections.}
285 :     chunk.sep <- "\n\n"
286 :     }
287 :     chunk.res <- NULL
288 : tdhock 191 if ( !grepl("^\\s*$",payload,perl=TRUE) )
289 : tdhock 87 chunk.res <-
290 :     if ( is.null(res[[field]]) ) payload
291 :     else paste(res[[field]], payload, sep=chunk.sep)
292 :     invisible(chunk.res)
293 :     }
294 : tdhock 154 while ( k <= length(src) ){
295 :     line <- src[k]
296 : tdhock 191 if ( grepl(extra.regexp,line,perl=TRUE) ){
297 : tdhock 87 ## we have a new extra chunk - first get field name and any payload
298 :     new.field <- gsub(extra.regexp,"\\1",line,perl=TRUE)
299 :     new.contents <- gsub(extra.regexp,"\\2",line,perl=TRUE)
300 :    
301 :     ##details<< As a special case, the construct \code{##describe<<} causes
302 :     ## similar processing to the main function arguments to be
303 :     ## applied in order to construct a describe block within the
304 :     ## documentation, for example to describe the members of a
305 :     ## list. All subsequent "same line" \code{##<<} comments go into that
306 :     ## block until terminated by a subsequent \code{##}\emph{xxx}\code{<<} line.
307 :     if ( "describe" == new.field ){
308 :     ##details<< Such regions may be nested, but not in such a way
309 :     ## that the first element in a \code{describe} is another \code{describe}.
310 :     ## Thus there must be at least one \code{##<<} comment between each
311 :     ## pair of \code{##describe<<} comments.
312 :     if ( first.describe ){
313 :     stop("consecutive ##describe<< at line",k,"in",name.fun)
314 :     } else {
315 :     if ( nzchar(new.contents) ){
316 :     if ( is.null(payload) || 0 == nzchar(payload) ){
317 :     payload <- new.contents
318 :     } else {
319 :     payload <- paste(payload,new.contents,sep="\n\n")
320 :     }
321 :     }
322 :     first.describe <- TRUE
323 :     }
324 :     } else if ( "end" == new.field ){
325 :     ##details<< When nested \code{describe} blocks are used, a comment-only
326 :     ## line with \code{##end<<} terminates the current level only; any
327 :     ## other valid \code{##}\emph{xxx}\code{<<} line terminates
328 :     ## all open describe blocks.
329 :     if ( in.describe>0 ){
330 :     ## terminate current \item and \describe block only
331 :     if ( "value" == cur.field && 1 == in.describe ){
332 :     payload <- paste(payload,"}",sep="")
333 :     } else {
334 :     payload <- paste(payload,"}\n}",sep="")
335 :     }
336 :     in.describe <- in.describe-1;
337 :     } else {
338 :     warning("mismatched ##end<< at line ",k," in ",name.fun)
339 :     }
340 :     if ( nzchar(new.contents) ){
341 :     if ( nzchar(payload) ){
342 :     payload <- paste(payload,new.contents,sep="\n")
343 :     } else {
344 :     payload <- new.contents
345 :     }
346 :     }
347 :     } else {
348 :     ## terminate all open \describe blocks (+1 because of open item)
349 :     if ( 0 < in.describe ){
350 :     if ( "value" != cur.field ){ # value is implicit describe block
351 :     payload <- paste(payload,"}",sep="")
352 :     }
353 :     while ( in.describe>0 ){
354 :     payload <- paste(payload,"}",sep="\n")
355 :     in.describe <- in.describe-1;
356 :     }
357 :     }
358 :     ## finishing any existing payload
359 :     if ( in.chunk ) res[[cur.field]] <- end.chunk(cur.field,payload)
360 :     in.chunk <- TRUE
361 :     cur.field <- new.field
362 :     payload <- new.contents
363 :     ##note<< The "value" section of a .Rd file is implicitly a describe
364 :     ## block and \code{##}\code{value}\code{<<} acts accordingly. Therefore
365 :     ## it automatically enables the describe block itemization (##<< after
366 :     ## list entries).
367 :     if ( "value" == new.field ){
368 :     first.describe <- TRUE;
369 :     }
370 :     }
371 : tdhock 191 } else if ( in.chunk && grepl(cont.re,line,perl=TRUE) ){
372 : tdhock 87 ## append this line to current chunk
373 : tdhock 191 if ( !grepl(prefix,line,perl=TRUE) ){
374 : tdhock 87 ##describe<< Any lines with "\code{### }" at the left hand
375 :     ## margin within the included chunks are handled separately,
376 :     ## so if they appear in the documentation they will appear
377 :     ## before the \code{##}\emph{xxx}\code{<}\code{<} chunks.
378 :     ### This one should not appear.
379 :     stripped <- gsub(cont.re,"",line,perl=TRUE)
380 :     if ( nzchar(payload) ){
381 :     payload <- paste(payload,stripped,sep="\n")
382 :     } else {
383 :     payload <- stripped
384 :     }
385 :     }
386 : tdhock 191 } else if ( grepl(arg.pat,line,perl=TRUE) ){
387 : tdhock 87 not.describe <- (0==in.describe && !first.describe)
388 :     if ( in.chunk && not.describe){
389 :     res[[cur.field]] <- end.chunk(cur.field,payload)
390 :     }
391 :     comment <- gsub(arg.pat,"\\3",line,perl=TRUE);
392 :     arg <- gsub(arg.pat,"\\\\item\\{\\1\\}",line,perl=TRUE)
393 :     in.chunk <- TRUE
394 :     if ( not.describe ){
395 : tdhock 115 ## TDH 2010-06-18 For item{}s in the documentation list names,
396 :     ## we don't need to have a backslash before, so delete it.
397 :     arg <- gsub("^[\\]+","",arg)
398 : tdhock 87 cur.field <- gsub("...","\\dots",arg,fix=TRUE) ##special case for dots
399 :     payload <- comment
400 :     } else {
401 :     ## this is a describe block, so we need to paste with existing
402 :     ## payload as a new \item.
403 :     if ( first.describe ){
404 :     ## for first item, need to add describe block starter
405 :     if ( "value" == cur.field ){
406 :     payload <- paste(payload,"\n",arg,"{",sep="")
407 :     } else {
408 :     payload <- paste(payload,"\\describe{\n",arg,"{",sep="")
409 :     }
410 :     first.describe <- FALSE
411 :     in.describe <- in.describe+1
412 :     } else {
413 :     ## subsequent item - terminate existing and start new
414 :     payload <- paste(payload,"}\n",arg,"{",sep="")
415 :     }
416 :     if ( nzchar(comment) ){
417 :     payload <- paste(payload,comment,sep="")
418 :     }
419 :     }
420 :     } else if ( in.chunk ){
421 :     if ( 0 == in.describe && !first.describe ){
422 :     ## reached an end to current field, but need to wait if in.describe
423 :     res[[cur.field]] <- end.chunk(cur.field,payload)
424 :     in.chunk <- FALSE
425 :     cur.field <- NULL
426 :     payload <- NULL
427 :     }
428 :     }
429 :     k <- k+1
430 :     }
431 :     ## finishing any existing payload
432 :     if ( 0 < in.describe ){
433 :     if ( "value" != cur.field ){ # value is implicit describe block
434 :     payload <- paste(payload,"}",sep="")
435 :     }
436 :     while ( in.describe>0 ){
437 :     payload <- paste(payload,"}",sep="\n")
438 :     in.describe <- in.describe-1;
439 :     }
440 :     }
441 :     if ( in.chunk ) res[[cur.field]] <- end.chunk(cur.field,payload)
442 :     res
443 :     ### Named list of character strings extracted from comments. For each
444 :     ### name N we will look for N\{...\} in the Rd file and replace it
445 :     ### with the string in this list (implemented in modify.Rd.file).
446 :     }
447 :    
448 : kmpont 212 leadingS3generic <- function # check whether function name is an S3 generic
449 :     ### Determines whether a function name looks like an S3 generic function
450 :     (name, ##<< name of function
451 :     env, ##<< environment to search for additional generics
452 :     ...) ##<< ignored here
453 :     {
454 :     ##details<< This function is one of the default parsers, but exposed as
455 :     ## possibly of more general interest. Given a function name of the form
456 :     ## x.y.z it looks for the generic function x applying to objects of class
457 :     ## y.z and also for generic function x.y applying to objects of class z.
458 :     ##
459 :     parts <- strsplit(name, ".", fixed = TRUE)[[1]]
460 :     l <- length(parts)
461 :     if (l > 1) {
462 :     for (i in 1:(l - 1)) {
463 :     ## Look for a generic function (known by the system or defined
464 :     ## in the package) that matches that part of the function name
465 :     generic <- paste(parts[1:i], collapse = ".")
466 :     if (any(generic %in% utils:::getKnownS3generics()) ||
467 :     utils:::findGeneric(generic, env) != "") {
468 :     object <- paste(parts[(i + 1):l], collapse = ".")
469 :     ##details<< Assumes that the first name which matches any known
470 :     ## generics is the target generic function, so if both x and x.y
471 :     ## are generic functions, will assume generic x applying to objects
472 :     ## of class y.z
473 :     ##value<< If a matching generic found returns a list with a single component:
474 :     return(list(.s3method=c(generic, object))) ##<< a character vector containing generic name and object name.
475 :     }
476 :     }
477 :     }
478 :     ##value<< If no matching generic functions are found, returns an empty list.
479 :     list()
480 :     }
481 :    
482 : tdhock 154 ### Parsers for each function that are constructed automatically. This
483 :     ### is a named list, and each element is a parser function for an
484 :     ### individual object.
485 :     forfun.parsers <-
486 :     list(prefixed.lines=prefixed.lines,
487 :     examples.after.return=examples.after.return,
488 :     extract.xxx.chunks=extract.xxx.chunks,
489 :     ## title from first line of function def
490 : tdhock 191 title.from.firstline=function(src,...){
491 : tdhock 195 first <- src[1]
492 :     if(!is.character(first))return(list())
493 :     if(!grepl("#",first))return(list())
494 :     list(title=gsub("[^#]*#\\s*(.*)","\\1",first,perl=TRUE))
495 : tdhock 154 },
496 :     ## PhG: it is tests/FUN.R!!! I would like more flexibility here
497 :     ## please, let me choose which dir to use for examples!
498 :     ## Get examples for FUN from the file tests/FUN.R
499 :     examples.from.testfile=function(name,...){
500 :     tsubdir <- getOption("inlinedocs.exdir")
501 :     if (is.null(tsubdir)) tsubdir <- "tests" # Default value
502 :     tfile <- file.path("..",tsubdir,paste(name,".R",sep=""))
503 :     if(file.exists(tfile))
504 : tdhock 185 list(examples=readLines(tfile))
505 : tdhock 154 else list()
506 :     },
507 :     definition.from.source=function(doc,src,...){
508 :     def <- doc$definition
509 :     is.empty <- function(x)is.null(x)||x==""
510 :     if(is.empty(def) && !is.empty(src))
511 : tdhock 185 list(definition=src)
512 : tdhock 154 else list()
513 :     })
514 :    
515 :     ### List of Parser Functions that can be applied to any object.
516 :     forall.parsers <-
517 :     list(## Fill in author from DESCRIPTION and titles.
518 :     author.from.description=function(desc,...){
519 :     list(author=desc[,"Maintainer"])
520 :     },
521 :     ## The format section sometimes causes problems, so erase it.
522 :     erase.format=function(...){
523 :     list(format="")
524 :     },
525 :     ## Convert the function name to a title.
526 :     title.from.name=function(name,doc,...){
527 :     if("title"%in%names(doc))list() else
528 :     list(title=gsub("[._]"," ",name))
529 :     },
530 :     ## PhG: here is what I propose for examples code in the 'ex' attribute
531 :     examples.in.attr = function (name, o, ...) {
532 :     ex <- attr(o, "ex")
533 :     if (!is.null(ex)) {
534 :     ## Special case for code contained in a function
535 :     if (inherits(ex, "function")) {
536 :     ## If source is available, start from there
537 :     src <- attr(ex, "source")
538 :     if (!is.null(src)) {
539 :     ex <- src
540 :     } else { ## Use the body of the function
541 :     ex <- deparse(body(ex))
542 :     }
543 :     ## Eliminate leading and trailing code
544 :     ex <- ex[-c(1, length(ex))]
545 : tdhock 163 ## all the prefixes
546 :     ex <- kill.prefix.whitespace(ex)
547 : tdhock 154 ## Add an empty line before and after example
548 :     ex <- c("", ex, "")
549 :     }
550 : tdhock 185 list(examples = ex)
551 : tdhock 154 } else list()
552 : tdhock 200 },collapse=function(doc,...){
553 :     L <- lapply(doc,paste,collapse="\n")
554 :     L$.overwrite <- TRUE
555 :     L
556 : kmpont 212 },tag.s3methods=leadingS3generic
557 : tdhock 196 )
558 : tdhock 154
559 :     ### List of parser functions that operate on single objects. This list
560 :     ### is useful for testing these functions.
561 : tdhock 196 lonely <- structure(c(forall.parsers,forfun.parsers),ex=function(){
562 : tdhock 154 f <- function # title
563 :     ### description
564 :     (x, ##<< arg x
565 :     y
566 :     ### arg y
567 :     ){
568 :     ##value<< a list with elements
569 :     list(x=x, ##<< original x value
570 :     y=y, ##<< original y value
571 :     sum=x+y) ##<< their sum
572 :     ##end<<
573 :     }
574 :     src <- attr(f,"source")
575 :     lonely$extract.xxx.chunks(src)
576 :     lonely$prefixed.lines(src)
577 : tdhock 196 })
578 : tdhock 154
579 :     extra.code.docs <- function # Extract documentation from code chunks
580 :     ### Parse R code to extract inline documentation from comments around
581 :     ### each function. These are not able to be retreived simply by
582 :     ### looking at the "source" attribute. This is a Parser Function that
583 :     ### can be used in the parser list of package.skeleton.dx(). TODO:
584 :     ### Modularize this into separate Parsers Functions for S4 classes,
585 :     ### prefixes, ##<<blocks, etc. Right now it is not very clean!
586 :     (code,
587 :     ### Code lines in a character vector containing multiple R objects to
588 :     ### parse for documentation.
589 :     objs,
590 :     ### The objects defined in the code.
591 :     ...
592 :     ### ignored
593 :     ){
594 :     parsed <- extract.file.parse(code)
595 :     extract.docs.try <- function(o,on)
596 :     {
597 :     ## Note: we could use parsed information here too, but that
598 :     ## would produce different results for setMethodS3 etc.
599 :     doc <- list()
600 :     if ( !is.null(parsed[[on]]) ){
601 :     if ( !is.na(parsed[[on]]@code[1]) ){ # no code given for generics
602 : tdhock 185 doc$definition <- paste(parsed[[on]]@code)
603 : tdhock 154 }
604 :     if(!"description"%in%names(doc) && !is.na(parsed[[on]]@description) ){
605 :     doc$description <- parsed[[on]]@description
606 :     }
607 : kmpont 212 ## if ( "setMethodS3" == parsed[[on]]@created ){
608 :     ## gen <- leadingS3generic(on,topenv())
609 :     ## if ( 0 < length(gen) ){
610 :     ## doc$.s3method <- gen$.s3method
611 :     ## cat("S3method(",gen$.s3method[1],",",gen$.s3method[2],")\n",sep="")
612 :     ## }
613 :     ## }
614 : tdhock 154 }
615 :     if("title" %in% names(doc) && !"description" %in% names(doc) ){
616 :     ## For short functions having both would duplicate, but a
617 :     ## description is required. Therefore automatically copy title
618 :     ## across to avoid errors at package build time.
619 :     doc$description <- doc$title
620 :     }
621 :     doc
622 :     }
623 :     extract.docs <- function(on){
624 :     res <- try({o <- objs[[on]]
625 :     extract.docs.try(o, on)},FALSE)
626 :     if(class(res)=="try-error"){
627 :     cat("Failed to extract docs for: ",on,"\n\n")
628 :     list()
629 :     } else if(0 == length(res) && inherits(objs[[on]],"standardGeneric")){
630 :     NULL
631 :     } else if(0 == length(res) && "function" %in% class(o)
632 :     && 1 == length(osource <- attr(o,"source"))
633 : tdhock 191 && grepl(paste("UseMethod(",on,")",sep="\""),osource)
634 : tdhock 154 ){
635 :     ## phew - this should only pick up R.oo S3 generic definitions like:
636 :     ## attr(*, "source")= chr "function(...) UseMethod(\"select\")"
637 :     NULL
638 :     } else res
639 :     }
640 :     doc.names <- names(objs)
641 :     res <- sapply(doc.names,extract.docs,simplify=FALSE)
642 :     ## Special processing for S4 classes as they do not appear in normal ls()
643 :     for ( nn in names(parsed) ){
644 :     if ( parsed[[nn]]@created == "setClass" ){
645 :     S4class.docs <- extract.docs.setClass(parsed[[nn]])
646 :     docname <- paste(nn,"class",sep="-")
647 :     if ( is.null(res[[docname]]) ){
648 :     res[[docname]] <- S4class.docs
649 :     doc.names <- c(doc.names,docname)
650 :     } else {
651 :     stop(nn," appears as both S4 class and some other definition")
652 :     }
653 :     }
654 :     }
655 :     inherit.docs <- function(on){
656 :     in.res <- res[[on]]
657 :     if ( !is.null(parsed[[on]]) ){
658 :     for ( parent in parsed[[on]]@parent ){
659 :     if ( !is.na(parent) ){
660 :     if ( is.null(in.res) ){
661 :     in.res <- res[[parent]]
662 :     } else if ( parent %in% names(res) ){
663 :     parent.docs <- res[[parent]]
664 :     for ( nn in names(parent.docs) ){
665 :     if ( !nn %in% names(in.res) ){
666 :     in.res[[nn]] <- parent.docs[[nn]]
667 :     }
668 :     }
669 :     }
670 :     }
671 :     }
672 :     }
673 :     invisible(in.res)
674 :     }
675 :     all.done <- FALSE
676 :     while ( !all.done ){
677 :     res1 <- sapply(doc.names,inherit.docs,simplify=FALSE)
678 :     all.done <- identical(res1,res)
679 :     res <- res1
680 :     }
681 :     ## now strip out any generics (which have value NULL in res):
682 :     res.not.null <- sapply(res,function(x){!is.null(x)})
683 :     if ( 0 < length(res.not.null) && length(res.not.null) < length(res) ){
684 :     res <- res[res.not.null]
685 :     }
686 :     res
687 :     ### named list of lists, one for each object to document.
688 :     }
689 :    
690 :     ### List of parsers to use by default with package.skeleton.dx.
691 :     default.parsers <-
692 :     c(extra.code.docs=extra.code.docs, ## TODO: cleanup!
693 :     sapply(forfun.parsers,forfun),
694 :     edit.package.file=function(desc,...){
695 :     in.details <- setdiff(colnames(desc),"Description")
696 : tdhock 185 details <- sprintf("%s: \\tab %s\\cr",in.details,desc[,in.details])
697 : tdhock 154 L <-
698 :     list(list(title=desc[,"Title"],
699 :     description=desc[,"Description"],
700 : tdhock 195 `tabular{ll}`=details))
701 : tdhock 154 names(L) <- paste(desc[,"Package"],"-package",sep="")
702 :     L
703 : tdhock 197 },
704 :     sapply(forall.parsers,forall)
705 : tdhock 186 )
706 : tdhock 154
707 : tdhock 87 setClass("DocLink", # Link documentation among related functions
708 :     ### The \code{.DocLink} class provides the basis for hooking together
709 :     ### documentation of related classes/functions/objects. The aim is that
710 : kmpont 212 ### documentation sections missing from the child are inherited from
711 :     ### the parent class.
712 : tdhock 87 representation(name="character", ##<< name of object
713 :     created="character", ##<< how created
714 :     parent="character", ##<< parent class or NA
715 :     code="character", ##<< actual source lines
716 :     description="character") ##<< preceding description block
717 :     )
718 :    
719 :     extract.file.parse <- function # File content analysis
720 :     ### Using the base \code{\link{parse}} function, analyse the file to link
721 :     ### preceding "prefix" comments to each active chunk. Those comments form
722 :     ### the default description for that chunk. The analysis also looks for
723 :     ### S4 class "setClass" calls and R.oo setConstructorS3 and setMethodS3
724 :     ### calls in order to link the documentation of those properly.
725 :     (code
726 :     ### Lines of R source code in a character vector - note that any
727 :     ### nested \code{source} statements are \emph{ignored} when scanning
728 :     ### for class definitions.
729 :     ){
730 :     res <- list()
731 :     old.opt <- options(keep.source=TRUE)
732 :     parsed <- try(parse(text=code))
733 :     options(old.opt)
734 :     if ( inherits(parsed,"try-error") ){
735 :     stop("parse failed with error:\n",parsed)
736 :     }
737 :     chunks <- attr(parsed,"srcref")
738 :     last.end <- 0
739 :     for ( k in 1:length(parsed) ){
740 :     start <- chunks[[k]][1]
741 :     ##details<< If the definition chunk does not contain a
742 :     ## description, any immediately preceding sequence consecutive
743 :     ## "prefix" lines will be used instead.
744 :     default.description <- NULL
745 :     while ( start > last.end+1
746 : tdhock 191 && grepl(prefix,code[start-1],perl=TRUE) ){
747 : tdhock 87 start <- start-1
748 :     }
749 :     if ( start < chunks[[k]][1] ){
750 :     default.description <- decomment(code[start:(chunks[[k]][1]-1)])
751 :     } else {
752 :     default.description <- NA_character_;
753 :     }
754 :     ##details<< Class and method definitions can take several forms,
755 :     ## determined by expression type: \describe{
756 :     ## \item{assignment (<-)}{Ordinary assignment of value/function;}
757 :     ## \item{setClass}{Definition of S4 class;}
758 :     ## \item{setConstructorS3}{Definition of S3 class using R.oo package;}
759 :     ## \item{setMethodS3}{Definition of method for S3 class using R.oo package.}}
760 :     ## Additionally, the value may be a name of a function defined elsewhere,
761 :     ## in which case the documentation should be copied from that other definition.
762 :     ## This is handled using the concept of documentation links.
763 :     lang <- parsed[[k]]
764 :     chars <- as.character(lang)
765 :     expr.type <- chars[1]
766 :     parent <- NA_character_
767 :    
768 :     if ( expr.type == "<-" || expr.type == "setConstructorS3" || expr.type == "setClass" ){
769 :     object.name <- chars[2]
770 :     ## If the function definition is not embedded within the call, then
771 :     ## the parent is that function. Test whether the the third value
772 :     ## looks like a name and add it to parents if so.
773 : tdhock 191 if ( grepl("^[\\._\\w]+$",chars[3],perl=TRUE) ){
774 : tdhock 87 parent <- chars[3]
775 :     }
776 :     res[[object.name]] <- new("DocLink",name=object.name,
777 :     created=expr.type,
778 :     parent=parent,
779 :     code=paste(chunks[[k]],sep=""),
780 :     description=default.description)
781 :     } else if ( expr.type == "setMethodS3" ){
782 :     ##details<< The \code{setMethodS3} calls introduce additional
783 :     ## complexity: they will define an additional S3 generic (which
784 :     ## needs documentation to avoid warnings at package build time)
785 :     ## unless one already exists. This also is handled by "linking"
786 :     ## documentation. A previously unseen generic is linked to the
787 :     ## first defining instances, subsequent definitions of that generic
788 :     ## also link back to the first defining instance.
789 :     generic.name <- chars[2]
790 :     object.name <- paste(generic.name,chars[3],sep=".")
791 :     if ( is.null(res[[generic.name]]) ){
792 :     generic.desc <- paste("Generic method behind \\code{\\link{",object.name,"}}",sep="")
793 :     res[[generic.name]] <- new("DocLink",
794 :     name=generic.name,
795 :     created=expr.type,
796 :     parent=object.name,
797 :     code=NA_character_,
798 :     description=generic.desc)
799 :     } else {
800 :     parent <- res[[generic.name]]@parent
801 :     }
802 :     ## If the function definition is not embedded within the call, then
803 :     ## the parent is that function. Test whether the the fourth value
804 :     ## looks like a name and add it to parents if so.
805 : tdhock 191 if ( grepl("^[\\._\\w]+$",chars[4],perl=TRUE) ){
806 : tdhock 87 parent <- c(chars[4],parent)
807 :     }
808 :     res[[object.name]] <- new("DocLink",name=object.name,
809 :     created=expr.type,
810 :     parent=parent,
811 :     code=paste(chunks[[k]],sep=""),
812 :     description=default.description)
813 :     } else {
814 :     ## Not sure what to do with these yet. Need to deal with setMethod, setAs etc.
815 :     }
816 :     }
817 :     invisible(res)
818 :     ### Returns an invisible list of .DocLink objects.
819 :     }
820 :    
821 :     extract.docs.setClass <- function # S4 class inline documentation
822 :     ### Using the same conventions as for functions, definitions of S4 classes
823 :     ### in the form \code{setClass("classname",\dots)} are also located and
824 :     ### scanned for inline comments.
825 : tdhock 93 (doc.link
826 : tdhock 87 ### DocLink object as created by \code{\link{extract.file.parse}}.
827 :     ### Note that \code{source} statements are \emph{ignored} when scanning for
828 :     ### class definitions.
829 : tdhock 93 ){
830 : tdhock 87 chunk.source <- doc.link@code
831 :     ##details<<
832 :     ## Extraction of S4 class documentation is currently limited to expressions
833 :     ## within the source code which have first line starting with
834 :     ## \code{setClass("classname"}. These are located from the source file
835 :     ## (allowing also for white space around the \code{setClass} and \code{(}).
836 :     ## Note that \code{"classname"} must be a quoted character string;
837 :     ## expressions returning such a string are not matched.
838 :     class.name <- doc.link@name
839 :    
840 :     ##details<< For class definitions, the slots (elements of the
841 :     ## \code{representation} list) fill the role of function
842 :     ## arguments, so may be documented by \code{##<<} comments on
843 :     ## the same line or \code{### } comments at the beginning of the
844 :     ## following line.
845 :     f.n <- paste(class.name,"class",sep="-")
846 : tdhock 154 docs <- extract.xxx.chunks(chunk.source,f.n)
847 : tdhock 138 ## also apply source parsing functions that I separated out into
848 :     ## separate functions
849 :     docs <- combine(docs,lonely$prefixed.lines(chunk.source))
850 :     docs$title <- lonely$title.from.firstline(chunk.source)
851 : tdhock 87 ##details<<
852 :     ## The class definition skeleton includes an \code{Objects from the Class}
853 :     ## section, to which any \code{##details<<} documentation chunks are
854 :     ## written. It is given a vanilla content if there are no specific
855 :     ## \code{##details<<} documentation chunks.
856 :     if ( is.null(docs[["details"]]) ){
857 :     docs[["details"]] <-
858 :     paste("Objects can be created by calls of the form \\code{new(",
859 :     class.name," ...)}",sep="")
860 :     }
861 :     docs[["section{Objects from the Class}"]] <- docs[["details"]]
862 :     ## seealso has a skeleton line not marked by ~ .. ~, so have to suppress
863 :     if ( is.null(docs[["seealso"]]) ){
864 :     docs[["seealso"]] <- ""
865 :     }
866 :     if ( is.null(docs[["alias"]]) ){
867 :     docs[["alias"]] <- class.name
868 :     }
869 :     if ( is.null(docs[["description"]]) ){
870 :     docs[["description"]] <- doc.link@description
871 :     }
872 :     invisible(docs)
873 :     }
874 : tdhock 105
875 : tdhock 109 apply.parsers <- function
876 :     ### Parse code to r objs, then run all the parsers and return the
877 :     ### documentation list.
878 : tdhock 105 (code,
879 :     ### Character vector of code lines.
880 : tdhock 109 parsers=default.parsers,
881 : tdhock 105 ### List of Parser Functions.
882 : tdhock 109 verbose=FALSE,
883 : tdhock 105 ### Echo names of Parser Functions?
884 :     ...
885 :     ### Additional arguments to pass to Parser Functions.
886 :     ){
887 :     e <- new.env()
888 : kmpont 212 ## KMP 2011-03-09 fix problem with DocLink when inlinedocs ran on itself
889 :     ## Error in assignClassDef(Class, classDef, where) :
890 :     ## Class "DocLink" has a locked definition in package "inlinedocs"
891 :     ## Traced to "where" argument in setClassDef which defaults to topenv()
892 :     ## which in turn is inlinedocs when processing inlinedocs package, hence
893 :     ## the clash. The following works (under R 2.12.2), so that the topenv()
894 :     ## now finds e before finding the inlinedocs environment.
895 : kmpont 213 old <- options(keep.source=TRUE,topLevelEnvironment=e)
896 : tdhock 109 on.exit(options(old))
897 : tdhock 110 exprs <- parse(text=code)
898 : tdhock 248 ## TDH 2011-04-07 set this so that no warnings about creating a fake
899 :     ## package when we try to process S4 classes defined in code
900 :     e$.packageName <- "inlinedocs.processor"
901 : tdhock 152 for (i in exprs){
902 : tdhock 248 ## TDH 2011-04-07 Disable this tryCatch since it makes it harder
903 :     ##to debug errors/warnings in the evaluated code
904 :    
905 :     ##tryCatch({
906 :     eval(i, e)
907 :     ##},error=function(e){
908 : tdhock 152 ##print(e)
909 : tdhock 248 ##})
910 : tdhock 152 }
911 : tdhock 105 objs <- sapply(ls(e),get,e,simplify=FALSE)
912 :    
913 :     docs <- list()
914 : phgrosjean 108
915 : tdhock 109 ## apply parsers in sequence to code and objs
916 : tdhock 105 for(i in seq_along(parsers)){
917 :     N <- names(parsers[i])
918 :     if(verbose){
919 :     if(is.character(N) && N!=""){
920 :     cat(N," ",sep="")
921 :     }else cat('. ')
922 :     }
923 :     p <- parsers[[i]]
924 :     ## This is the argument list that each parser receives:
925 : tdhock 200 L <- p(code=code,objs=objs,docs=docs,env=e,...)
926 : tdhock 105 docs <- combine(docs,L)
927 :     }
928 : tdhock 196 ## post-process to collapse all character vectors
929 :     for(i in seq_along(docs)){
930 :     for(j in seq_along(docs[[i]])){
931 : tdhock 198 if(names(docs[[i]])[j]!=".s3method")
932 : tdhock 196 docs[[i]][[j]] <- paste(docs[[i]][[j]],collapse="\n")
933 :     }
934 :     }
935 : tdhock 105 if(verbose)cat("\n")
936 :     docs
937 :     ### A list of extracted documentation from code.
938 :     }
939 : tdhock 109
940 : tdhock 138 ### Names of Parser Functions that operate on the desc arg.
941 :     descfile.names <- c("author.from.description","edit.package.file")
942 :    
943 :     ### Names of Parser Functions that do NOT use the desc arg.
944 :     non.descfile.names <-
945 :     names(default.parsers)[!names(default.parsers)%in%descfile.names]
946 :    
947 : tdhock 118 ### Parsers that operate only on R code, independently of the
948 :     ### description file.
949 : tdhock 138 nondesc.parsers <- default.parsers[non.descfile.names]
950 : tdhock 118
951 : tdhock 209 extract.docs.file <- structure(function
952 : tdhock 109 ### Apply all parsers relevant to extract info from just 1 code file.
953 :     (f,
954 :     ### File name of R code to read and parse.
955 : tdhock 203 parsers=NULL,
956 : tdhock 109 ### Parser Functions to use to parse the code and extract
957 :     ### documentation.
958 :     ...
959 :     ### Other arguments to pass to Parser Functions.
960 :     ){
961 : tdhock 203 if(is.null(parsers))parsers <- nondesc.parsers
962 : tdhock 109 apply.parsers(readLines(f),parsers,verbose=FALSE,...)
963 : tdhock 209 },ex=function(){
964 :     f <- system.file("silly","R","silly.R",package="inlinedocs")
965 : tdhock 248 extract.docs.file(f)
966 : tdhock 209 })
967 : tdhock 248

R-Forge@R-project.org
ViewVC Help
Powered by ViewVC 1.0.0  
Thanks to:
Vienna University of Economics and Business University of Wisconsin - Madison Powered By FusionForge