summaryrefslogtreecommitdiffstats
path: root/Doc/tools/buildindex.py
blob: 245122133b759e107581c6c91cbda5b9a4aaeabc (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
#! /usr/bin/env python

__version__ = '$Revision$'

import os
import re
import string
import sys


class Node:
    __rmjunk = re.compile("<#\d+#>")

    continuation = 0

    def __init__(self, link, str, seqno):
        self.links = [link]
        self.seqno = seqno
        # remove <#\d+#> left in by moving the data out of LaTeX2HTML
        str = self.__rmjunk.sub('', str)
        # build up the text
        self.text = split_entry_text(str)
        self.key = split_entry_key(str)

    def __cmp__(self, other):
        """Comparison operator includes sequence number, for use with
        list.sort()."""
        return self.cmp_entry(other) or cmp(self.seqno, other.seqno)

    def cmp_entry(self, other):
        """Comparison 'operator' that ignores sequence number."""
        c = 0
        for i in range(min(len(self.key), len(other.key))):
            c = (cmp_part(self.key[i], other.key[i])
                 or cmp_part(self.text[i], other.text[i]))
            if c:
                break
        return c or cmp(self.key, other.key) or cmp(self.text, other.text)

    def __repr__(self):
        return "<Node for %s (%s)>" % (string.join(self.text, '!'), self.seqno)

    def __str__(self):
        return string.join(self.key, '!')

    def dump(self):
        return "%s\1%s###%s\n" \
               % (string.join(self.links, "\1"),
                  string.join(self.text, '!'),
                  self.seqno)


def cmp_part(s1, s2):
    result = cmp(s1, s2)
    if result == 0:
        return 0
    l1 = string.lower(s1)
    l2 = string.lower(s2)
    minlen = min(len(s1), len(s2))
    if len(s1) < len(s2) and l1 == l2[:len(s1)]:
        result = -1
    elif len(s2) < len(s1) and l2 == l1[:len(s2)]:
        result = 1
    else:
        result = cmp(l1, l2) or cmp(s1, s2)
    return result


def split_entry(str, which):
    stuff = []
    parts = string.split(str, '!')
    parts = map(string.split, parts, ['@'] * len(parts))
    for entry in parts:
        if len(entry) != 1:
            key = entry[which]
        else:
            key = entry[0]
        stuff.append(key)
    return stuff


_rmtt = re.compile(r"""(.*)<tt(?: class=['"][a-z0-9]+["'])?>(.*)</tt>(.*)$""",
                   re.IGNORECASE)
_rmparens = re.compile(r"\(\)")

def split_entry_key(str):
    parts = split_entry(str, 1)
    for i in range(len(parts)):
        m = _rmtt.match(parts[i])
        if m:
            parts[i] = string.join(m.group(1, 2, 3), '')
        else:
            parts[i] = string.lower(parts[i])
        # remove '()' from the key:
        parts[i] = _rmparens.sub('', parts[i])
    return map(trim_ignored_letters, parts)


def split_entry_text(str):
    if '<' in str:
        m = _rmtt.match(str)
        if m:
            str = string.join(m.group(1, 2, 3), '')
    return split_entry(str, 1)


def load(fp):
    nodes = []
    rx = re.compile("(.*)\1(.*)###(.*)$")
    while 1:
        line = fp.readline()
        if not line:
            break
        m = rx.match(line)
        if m:
            link, str, seqno = m.group(1, 2, 3)
            nodes.append(Node(link, str, seqno))
    return nodes


def trim_ignored_letters(s):
    # ignore $ to keep environment variables with the
    # leading letter from the name
    s = string.lower(s)
    if s[0] == "$":
        return s[1:]
    else:
        return s

def get_first_letter(s):
    return string.lower(trim_ignored_letters(s)[0])


def split_letters(nodes):
    letter_groups = []
    if nodes:
        group = []
        append = group.append
        letter = get_first_letter(nodes[0].text[0])
        letter_groups.append((letter, group))
        for node in nodes:
            nletter = get_first_letter(node.text[0])
            if letter != nletter:
                letter = nletter
                group = []
                letter_groups.append((letter, group))
                append = group.append
            append(node)
    return letter_groups


# need a function to separate the nodes into columns...
def split_columns(nodes, columns=1):
    if columns <= 1:
        return [nodes]
    # This is a rough height; we may have to increase to avoid breaks before
    # a subitem.
    colheight = len(nodes) / columns
    numlong = len(nodes) % columns
    if numlong:
        colheight = colheight + 1
    else:
        numlong = columns
    cols = []
    for i in range(numlong):
        start = i * colheight
        end = start + colheight
        cols.append(nodes[start:end])
    del nodes[:end]
    colheight = colheight - 1
    try:
        numshort = len(nodes) / colheight
    except ZeroDivisionError:
        cols = cols + (columns - len(cols)) * [[]]
    else:
        for i in range(numshort):
            start = i * colheight
            end = start + colheight
            cols.append(nodes[start:end])
    #
    # If items continue across columns, make sure they are marked
    # as continuations so the user knows to look at the previous column.
    #
    for i in range(len(cols) - 1):
        try:
            prev = cols[i][-1]
            next = cols[i + 1][0]
        except IndexError:
            return cols
        else:
            n = min(len(prev.key), len(next.key))
            for j in range(n):
                if prev.key[j] != next.key[j]:
                    break
                next.continuation = j + 1
    return cols


DL_LEVEL_INDENT = "  "

def format_column(nodes):
    strings = ["<dl compact>"]
    append = strings.append
    level = 0
    previous = []
    for node in nodes:
        current = node.text
        count = 0
        for i in range(min(len(current), len(previous))):
            if previous[i] != current[i]:
                break
            count = i + 1
        if count > level:
            append("<dl compact>" * (count - level) + "\n")
            level = count
        elif level > count:
            append("\n")
            append(level * DL_LEVEL_INDENT)
            append("</dl>" * (level - count))
            level = count
        # else: level == count
        for i in range(count, len(current) - 1):
            term = node.text[i]
            level = level + 1
            if node.continuation > i:
                extra = " (continued)"
            else:
                extra = ""
            append("\n<dt>%s%s\n<dd>\n%s<dl compact>"
                   % (term, extra, level * DL_LEVEL_INDENT))
        append("\n%s<dt>%s%s</a>"
               % (level * DL_LEVEL_INDENT, node.links[0], node.text[-1]))
        for link in node.links[1:]:
            append(",\n%s    %s[Link]</a>" % (level * DL_LEVEL_INDENT, link))
        previous = current
    append("\n")
    append("</dl>" * (level + 1))
    return string.join(strings, '')


def format_nodes(nodes, columns=1):
    strings = []
    append = strings.append
    if columns > 1:
        colnos = range(columns)
        colheight = len(nodes) / columns
        if len(nodes) % columns:
            colheight = colheight + 1
        colwidth = 100 / columns
        append('<table width="100%"><tr valign="top">')
        for col in split_columns(nodes, columns):
            append('<td width="%d%%">\n' % colwidth)
            append(format_column(col))
            append("\n</td>")
        append("\n</tr></table>")
    else:
        append(format_column(nodes))
    append("\n<p>\n")
    return string.join(strings, '')


def format_letter(letter):
    if letter == '.':
        lettername = ". (dot)"
    elif letter == '_':
        lettername = "_ (underscore)"
    else:
        lettername = string.upper(letter)
    return "\n<hr>\n<h2><a name=\"letter-%s\">%s</a></h2>\n\n" \
           % (letter, lettername)


def format_html_letters(nodes, columns=1):
    letter_groups = split_letters(nodes)
    items = []
    for letter, nodes in letter_groups:
        s = "<b><a href=\"#letter-%s\">%s</a></b>" % (letter, letter)
        items.append(s)
    s = ["<hr><center>\n%s</center>\n" % string.join(items, " |\n")]
    for letter, nodes in letter_groups:
        s.append(format_letter(letter))
        s.append(format_nodes(nodes, columns))
    return string.join(s, '')

def format_html(nodes, columns):
    return format_nodes(nodes, columns)


def collapse(nodes):
    """Collapse sequences of nodes with matching keys into a single node.
    Destructive."""
    if len(nodes) < 2:
        return
    prev = nodes[0]
    i = 1
    while i < len(nodes):
        node = nodes[i]
        if not node.cmp_entry(prev):
            prev.links.append(node.links[0])
            del nodes[i]
        else:
            i = i + 1
            prev = node


def dump(nodes, fp):
    for node in nodes:
        fp.write(node.dump())


def process_nodes(nodes, columns, letters):
    nodes.sort()
    collapse(nodes)
    if letters:
        return format_html_letters(nodes, columns)
    else:
        return format_html(nodes, columns)


def main():
    import getopt
    ifn = "-"
    ofn = "-"
    columns = 1
    letters = 0
    opts, args = getopt.getopt(sys.argv[1:], "c:lo:",
                               ["columns=", "letters", "output="])
    for opt, val in opts:
        if opt in ("-o", "--output"):
            ofn = val
        elif opt in ("-c", "--columns"):
            columns = string.atoi(val)
        elif opt in ("-l", "--letters"):
            letters = 1
    if not args:
        args = [ifn]
    nodes = []
    for fn in args:
        nodes = nodes + load(open(fn))
    num_nodes = len(nodes)
    html = process_nodes(nodes, columns, letters)
    program = os.path.basename(sys.argv[0])
    if ofn == "-":
        sys.stdout.write(html)
        sys.stderr.write("\n%s: %d index nodes" % (program, num_nodes))
    else:
        open(ofn, "w").write(html)
        print
        print "%s: %d index nodes" % (program, num_nodes)


if __name__ == "__main__":
    main()