转帖老文:sed in (pseudo) microcode

waker

sed in (pseudo) microcode, by Carlos J. Duarte.

1. 
   
2. sed in (pseudo) microcode, by Carlos J. Duarte.
   
3. [ This article was posted to the seders mailing list ]
   
4. 
   
5. Al, this is a little doc, with some pseudo code, of
   
6. a sed implementation.
   
7. 
   
8. The idea, is when people (that knows sed), has a question
   
9. or need some fine tunning, to quickly answered themselves.
   
10. 
    
11. One particulary example, are the strange behaviors of D, and a\
    
12. (well, sort of).
    
13. 
    
14. As a last note remark, I am revieing my scripts packages, and
    
15. also some documentation on sed (do-it-with-sed, advanced-sed,
    
16. hints-sed? or sed-fine-tunning). This should be ok on august.
    
17. For now (upto end of july), I will not (probably), dedicate
    
18. any time to that.
    
19. 
    
20. So, there goes the text.
    
21. 
    
22. ADDRESSES
    
23. =========
    
24. 
    
25. (0) no address, just the command per se
    
26. (1) a single address or none, is permitted
    
27. (2) a single address, a double address, or none is permitted
    
28. 
    
29. any address, if specified, may be a number -- specifies
    
30. the input line number, or a regular expression -- specifies
    
31. all pattern spaces, that matches it.
    
32. 
    
33. one address, appears per se, before the command
    
34. two addresses, appears separated per a comma `,'
    
35. 
    
36. if an address is omitted, the command will be apllied
    
37. on all pattern spaces that passes through it
    
38. 
    
39. y/a/b/                  executes on all ps
    
40. 3 y/a/b/                executes on ps which last line loaded, was 3
    
41. /foo/ y/a/b/            executes on ps that matches /foo/
    
42. 4,/foo/ y/a/b/          executes from ps which last line loaded was 4,
    
43.                         up to, including, ps that matches /foo/
    
44. /foo/,5 y/a/b/          as above, but from /foo/ up to 5
    
45. 
    
46. COMMANDS MICROCODE
    
47. ==================
    
48. 
    
49. notes:
    
50. - first, everything begins with a (sic) `goto [end]', after
    
51.   all variables have been zeroed/offed/emptyed
    
52. 
    
53. - all commands, have a `pop ps' and a `push ps', to enforce that
    
54.   they really operate on patern space
    
55. 
    
56. - the `print ps to stdout' "microcode" below, prints an extra newline
    
57.   after the print the `ps' itself
    
58. 
    
59.         [end]
    
60.         . pop ps
    
61.         . if not nflag and line_counter>=1, print ps to stdout
    
62.         . print append_buffer to stdout (which might be empty)
    
63.         . empty append_buffer
    
64.         . set tflag off (see s, t)
    
65.         . if come_from_D
    
66.         .       set come_from_D off (see D)
    
67.         . else
    
68.         .       if no more input lines, exit
    
69.         .       load next line into ps, increment line_counter
    
70.         . push ps
    
71.         . goto start
    
72. 
    
73. (2) s/RE/replacement/[gp#w], #: 1-9
    
74.         . pop ps
    
75.         . if # not given, or g given, #=1
    
76.         . set last_match = begin of ps
    
77.         !! do not set tflag off (see t, [end])
    
78.         [1]
    
79.         . repeat # searches for RE, on ps, from last_match to end
    
80.         . if found
    
81.         .       tflag on
    
82.         .       replace the found match, per `replacement'
    
83.         .       if w file, append ps at end of `file'
    
84.         .       if p, print ps to stdout
    
85.         .       if g, set last_match to end of replacement+1 and goto [1]
    
86.         . push ps
    
87. 
    
88. (2) y/list1/list2/
    
89.         . pop ps
    
90.         . foreach character c1 on list1
    
91.         .       fetch character c2 from list2
    
92.         .       replace all occurrences of c1 per c2
    
93.         . push ps
    
94. (1)i\
    
95. text
    
96.         . print text to stdout immediatly
    
97. 
    
98. (1)a\
    
99. text
    
100.         . collect (i.e append) text to append_buffer (see [end])
     
101. 
     
102. (2)c\
     
103. text
     
104.         . print text to stdout immediatly
     
105.         . pop ps
     
106.         . empty ps
     
107.         . push ps
     
108.         . goto [end]
     
109.         !! because ps is empty, the `print ps' at [end] will not
     
110.            produce output
     
111. 
     
112. (1)r file
     
113.         . print contents of file, immediatly to stdout
     
114. 
     
115. (2)w file
     
116.         . pop ps
     
117.         . append ps at end of file
     
118.         . push ps
     
119. 
     
120. (0): label
     
121.         . name next command, as `label'
     
122. 
     
123. (2)b label
     
124.         . goto command named `label'
     
125. 
     
126. (2)t label
     
127.         . if tflag on (see s///)
     
128.         .       set tflag off
     
129.         .       goto command `label'
     
130. 
     
131. (1)q
     
132.         . pop ps
     
133.         . if not nflag, print ps to stdout
     
134.         . print append_buffer (might be empty, which cause no output)
     
135.         . exit
     
136. 
     
137. (2)p
     
138.         . pop ps
     
139.         . print ps to stdout
     
140.         . push ps
     
141. 
     
142. (2)l
     
143.         . pop ps
     
144.         . copy ps to tmp
     
145.         . replace on tmp, some special characters, per their
     
146.           conventional equivalents \x
     
147.         . print tmp
     
148.         . push ps
     
149. 
     
150. (1)=
     
151.         . print line_counter to stdout
     
152. 
     
153. (2)n
     
154.         . if no more lines on input, goto [end]
     
155.         . pop ps
     
156.         . if not nflag, print ps to stdout
     
157.         . load next line into ps, increment line_counter
     
158.         . push ps
     
159.         !! and continues, does not jump to start, or end
     
160. 
     
161. (2)d
     
162.         . pop ps
     
163.         . empty ps
     
164.         . push ps
     
165.         . goto [end]
     
166. 
     
167. (2)h
     
168.         . pop ps
     
169.         . set hold_buffer as ps
     
170.         . push ps
     
171. 
     
172. (2)H
     
173.         . pop ps
     
174.         . set hold_buffer to hold_buffer+\n+ps
     
175.         . push ps
     
176. 
     
177. (2)g
     
178.         . pop ps
     
179.         . set ps as hold_buffer
     
180.         . push ps
     
181. 
     
182. (2)G
     
183.         . pop ps
     
184.         . set ps as ps+\n+hold_buffer
     
185.         . push ps
     
186. 
     
187. (2)x
     
188.         . pop ps
     
189.         . push hold_buffer
     
190.         . set hold_buffer as ps
     
191.         !! exchange hold buffer with pattern space
     
192. 
     
193. (2)N
     
194.         . if no more lines on input, goto [end]
     
195.         . pop ps
     
196.         . load next line into tmp, increment line_counter
     
197.         . set ps to ps+\n+tmp
     
198.         . push ps
     
199.         !! and continues, does not jump to start, or end
     
200. 
     
201. (2)P
     
202.         . pop ps
     
203.         . copy ps to tmp
     
204.         . print remove from first \n upto end
     
205.         . print ps to stdout
     
206.         . push tmp
     
207. 
     
208. (2)D
     
209.         . pop ps
     
210.         . if ps does not contain a \n
     
211.         .       empty ps
     
212.         .       push ps
     
213.         .       goto [end]
     
214.         !! i.e. does the same as `d'
     
215.         . remove from beginning of ps, upto, including, first \n
     
216.         . push ps
     
217.         . set come_from_D on (see [end])
     
218.         . goto [end]
     
219.         !! do not load next line, if ps contained \n
     
220. 
     
221. ==
     
222. Carlos Duarte, 980712
     
223. 
     
224. 
     
225. 
     

复制代码

waker

最近有不少SED痴汉,可以把一些代码对照这个和sedsed的结果琢磨一下

li2002

mark 备查