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