Mercurial > hg > lxmldump
annotate lxmldump.py @ 26:420f13925f20
Rename a variable.
| author | Matti Hamalainen <ccr@tnsp.org> |
|---|---|
| date | Tue, 25 May 2021 11:39:36 +0300 |
| parents | 8a6738f67106 |
| children | d77ab8a300b1 |
| rev | line source |
|---|---|
| 0 | 1 #!/usr/bin/python3 -B |
| 2 # coding=utf-8 | |
| 3 ### | |
| 4 | 4 ### lxmldump - Dump ISO/FDIS 1951 XML file data |
| 5 ### Programmed and designed by Matti 'ccr' Hämäläinen <ccr@tnsp.org> | |
| 6 ### (C) Copyright 2021 Tecnic Software productions (TNSP) | |
| 7 ### | |
| 23 | 8 ### Released / distributed under 3-clause BSD license |
| 9 ### (see file "COPYING" for more information) | |
| 10 ### | |
| 4 | 11 ### Python 3.7+ required! |
| 0 | 12 ### |
| 13 import sys | |
| 14 import signal | |
| 15 import re | |
| 16 from pathlib import Path | |
| 17 import xml.etree.ElementTree as xmlET | |
|
5
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
18 import unicodedata |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
19 import argparse |
| 0 | 20 |
| 21 assert sys.version_info >= (3, 7) | |
| 22 | |
| 23 | |
| 24 ### | |
| 25 ### Default settings | |
| 26 ### | |
|
25
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
27 # Operation modes |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
28 PKK_MODE_NORMAL = 0 |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
29 PKK_MODE_DUMP = 1 |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
30 PKK_MODE_XML = 2 |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
31 |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
32 |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
33 # Element annotation mappings |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
34 pkk_element_annotation_map = { |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
35 "Fragment" : { |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
36 PKK_MODE_NORMAL: ["<", ">"], |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
37 }, |
| 7 | 38 } |
| 39 | |
| 40 | |
| 41 pkk_debug_list = [ | |
| 42 "ahas", | |
| 43 "ahavakkaine", | |
| 44 "ahavakala", | |
| 45 "ahavakoittuo", | |
| 46 "ahvaliha", | |
| 9 | 47 "aloilleh", |
| 18 | 48 "hanjahtoakseh", |
| 19 | 49 "akkalisto", |
| 7 | 50 ] |
| 51 | |
| 52 | |
| 21 | 53 |
| 0 | 54 ### |
| 55 ### Misc. helper functions, etc | |
| 56 ### | |
| 57 def pkk_cleanup(): | |
| 58 return 0 | |
| 59 | |
| 60 | |
| 13 | 61 ## Print string to stdout using normalized Unicode if enabled |
|
5
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
62 def pkk_print(smsg): |
| 12 | 63 try: |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
64 if pkk_cfg.normalize: |
| 12 | 65 sys.stdout.write(unicodedata.normalize("NFC", smsg)) |
| 66 else: | |
| 67 sys.stdout.write(smsg) | |
| 68 | |
| 69 except (BrokenPipeError, IOError) as e: | |
| 70 sys.stderr.close() | |
| 71 | |
| 0 | 72 |
| 13 | 73 ## Print string with indentation |
| 7 | 74 def pkk_printi(indent, smsg): |
| 22 | 75 pkk_print((" " * pkk_cfg.indent * indent) + smsg) |
| 7 | 76 |
| 0 | 77 |
| 78 ## Fatal error handler | |
| 79 def pkk_fatal(smsg): | |
| 80 print(u"ERROR: "+ smsg) | |
| 81 sys.exit(1) | |
| 82 | |
| 83 | |
| 84 ## Handler for SIGINT signals | |
| 85 def pkk_signal_handler(signal, frame): | |
| 86 pkk_cleanup() | |
| 87 print(u"\nQuitting due to SIGINT / Ctrl+C!") | |
| 88 sys.exit(1) | |
| 89 | |
| 90 | |
|
25
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
91 ## Annotate string |
| 26 | 92 def pkk_str_annotate(mtag, mstr): |
|
25
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
93 if pkk_cfg.annotate and mtag in pkk_element_annotation_map: |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
94 if pkk_cfg.mode in pkk_element_annotation_map[mtag]: |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
95 mmode = pkk_cfg.mode |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
96 else: |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
97 mmode = PKK_MODE_NORMAL |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
98 |
| 26 | 99 return pkk_element_annotation_map[mtag][mmode][0] + mstr + pkk_element_annotation_map[mtag][mmode][1] |
|
25
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
100 else: |
| 26 | 101 return mstr |
|
25
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
102 |
|
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
103 |
| 13 | 104 ## Clean string by removing tabs and newlines |
| 10 | 105 def pkk_str_clean(mstr): |
| 106 return re.sub(r'[\n\r\t]', '', mstr) | |
| 107 | |
| 108 | |
| 13 | 109 ## Format "Ptr" node as text |
| 9 | 110 def pkk_ptr_to_text(pnode): |
| 16 | 111 return "<PTR:{}>{}</PTR>".format( |
| 10 | 112 pnode.attrib["{http://www.w3.org/TR/xlink}href"], |
| 113 ("".join(pnode.itertext())).strip()) | |
| 9 | 114 |
| 115 | |
| 13 | 116 ## Get text inside a given node |
| 16 | 117 def pkk_node_to_text(lnode): |
| 7 | 118 stmp = "" |
| 119 for pnode in lnode.iter(): | |
| 9 | 120 if pnode.tag == "Ptr": |
| 121 stmp += pkk_ptr_to_text(pnode) | |
| 122 else: | |
| 123 if isinstance(pnode.text, str): | |
|
25
8a6738f67106
Factor string annotation into separate function.
Matti Hamalainen <ccr@tnsp.org>
parents:
24
diff
changeset
|
124 stmp += pkk_str_annotate(pnode.tag, pkk_str_clean(pnode.text).strip()) |
| 7 | 125 |
| 9 | 126 if isinstance(pnode.tail, str): |
| 10 | 127 stmp += pkk_str_clean(pnode.tail) |
| 7 | 128 |
| 129 return stmp.strip() | |
| 130 | |
| 131 | |
| 13 | 132 ## Simple recursive dump starting at given node |
| 7 | 133 def pkk_dump_recursive(indent, lnode): |
| 134 if lnode.tag in ["Example"]: | |
| 16 | 135 stmp = pkk_node_to_text(lnode) |
| 9 | 136 pkk_printi(indent, "{} \"{}\"\n".format(lnode.tag, stmp)) |
| 6 | 137 else: |
| 7 | 138 if isinstance(lnode.text, str): |
| 10 | 139 stmp = pkk_str_clean(lnode.text).strip() |
| 7 | 140 if stmp != "": |
| 141 stmp = " \""+ stmp +"\"" | |
| 142 else: | |
| 143 stmp = "" | |
| 6 | 144 |
| 145 if len(lnode.attrib) > 0: | |
| 10 | 146 atmp = " "+ str(lnode.attrib) |
| 6 | 147 else: |
| 148 atmp = "" | |
| 149 | |
| 7 | 150 pkk_printi(indent, "{}{}{}\n".format(lnode.tag, atmp, stmp)) |
| 6 | 151 for qnode in lnode.findall("./*"): |
| 7 | 152 pkk_dump_recursive(indent + 1, qnode) |
| 6 | 153 |
| 154 | |
| 17 | 155 ## Output item(s) under given node with given format string |
| 13 | 156 def pkk_output_subs_fmt(indent, dnode, dsub, dname, dfmt): |
| 7 | 157 for qnode in dnode.findall(dsub): |
| 16 | 158 pkk_printi(indent, dfmt.format(dname, pkk_node_to_text(qnode))) |
| 7 | 159 |
| 17 | 160 ## Output item(s) under given node with a prefixed name string |
| 13 | 161 def pkk_output_subs_prefix(indent, dnode, dsub, dname): |
| 162 pkk_output_subs_fmt(indent, dnode, dsub, dname, "{0} \"{1}\"\n") | |
| 7 | 163 |
| 164 | |
| 17 | 165 ## Output a main "Headword" or "Sense" node under it |
| 7 | 166 def pkk_output_sense(indent, dnode): |
| 17 | 167 # Search form and definition |
| 13 | 168 pkk_output_subs_prefix(indent, dnode, "./SearchForm", "srch") |
| 169 pkk_output_subs_prefix(indent, dnode, "./Definition", "defn") | |
| 7 | 170 |
| 17 | 171 # Examples |
| 7 | 172 for wnode in dnode.findall("./ExampleBlock/ExampleCtn"): |
| 16 | 173 sstr = pkk_node_to_text(wnode.find("./Example")) |
| 8 | 174 lstr = "" |
| 6 | 175 |
|
24
8c8e8e4504bb
Remove verbosity option as it is not really used.
Matti Hamalainen <ccr@tnsp.org>
parents:
23
diff
changeset
|
176 ltmp = [] |
|
8c8e8e4504bb
Remove verbosity option as it is not really used.
Matti Hamalainen <ccr@tnsp.org>
parents:
23
diff
changeset
|
177 for qnode in wnode.findall("./FreeTopic[@type='levikki']/GeographicalUsage"): |
|
8c8e8e4504bb
Remove verbosity option as it is not really used.
Matti Hamalainen <ccr@tnsp.org>
parents:
23
diff
changeset
|
178 ltmp.append("{} [{}]".format(pkk_node_to_text(qnode), qnode.attrib["class"])) |
| 8 | 179 |
|
24
8c8e8e4504bb
Remove verbosity option as it is not really used.
Matti Hamalainen <ccr@tnsp.org>
parents:
23
diff
changeset
|
180 if len(ltmp) > 0: |
|
8c8e8e4504bb
Remove verbosity option as it is not really used.
Matti Hamalainen <ccr@tnsp.org>
parents:
23
diff
changeset
|
181 lstr = " ({})".format(", ".join(ltmp)) |
| 7 | 182 |
| 183 pkk_printi(indent + 1, "{} \"{}\"{}\n".format("exmp", sstr, lstr)) | |
| 184 | |
| 185 | |
| 17 | 186 ## Output one "DictionaryEntry" node |
| 7 | 187 def pkk_output_node(indent, dnode): |
| 6 | 188 |
| 7 | 189 for wnode in dnode.findall("./HeadwordCtn"): |
| 18 | 190 # Create list with grammatical attributes (noun, verb, etc.) |
| 191 tmpl = [] | |
| 192 for pnode in wnode.findall("./PartOfSpeechCtn/PartOfSpeech"): | |
| 193 tmpl.append(pnode.attrib["freeValue"]) | |
| 194 | |
| 195 for pnode in wnode.findall("./GrammaticalNote"): | |
| 19 | 196 tmpl.append(pkk_node_to_text(pnode)) |
| 18 | 197 |
| 198 # Remove duplicates and sort the list | |
| 199 tmpl = list(set(tmpl)) | |
| 200 tmpl.sort(reverse=False, key=lambda attr: (attr, len(attr))) | |
| 201 | |
| 202 # Print the headword and attributes if any | |
| 203 pkk_output_subs_fmt(indent, wnode, "./Headword", "", "\"{1}\"") | |
| 204 if len(tmpl) > 0: | |
| 205 pkk_print(" ({})\n".format(" ; ".join(tmpl))) | |
| 206 else: | |
| 207 pkk_print("\n") | |
| 208 | |
| 209 # Print main "sense" | |
| 7 | 210 pkk_output_sense(indent + 1, wnode) |
| 6 | 211 |
| 18 | 212 # Print any other "senses" |
| 7 | 213 index = 1 |
| 214 for wnode in dnode.findall("./SenseGrp"): | |
| 8 | 215 pkk_printi(indent + 1, "sense #{}\n".format(index)) |
| 7 | 216 pkk_output_sense(indent + 2, wnode) |
| 217 index += 1 | |
| 6 | 218 |
| 219 | |
| 0 | 220 ### |
| 221 ### Main program starts | |
| 222 ### | |
| 223 signal.signal(signal.SIGINT, pkk_signal_handler) | |
| 224 | |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
225 optparser = argparse.ArgumentParser( |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
226 description="lxmldump - Dump ISO/FDIS 1951 XML file data", |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
227 usage="%(prog)s [options] <input xml file(s)>", |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
228 epilog="\n\n" |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
229 ) |
| 0 | 230 |
| 22 | 231 optparser.add_argument("filenames", |
| 232 type=str, action="extend", nargs="*", | |
| 233 metavar="filename", | |
| 234 help="XML filename(s)") | |
| 0 | 235 |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
236 optparser.add_argument("-d", "--dump", |
| 22 | 237 dest="mode", |
| 21 | 238 action="store_const", const=PKK_MODE_DUMP, default=PKK_MODE_NORMAL, |
| 22 | 239 help="output as simple dump") |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
240 |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
241 optparser.add_argument("-x", "--xml", |
| 22 | 242 dest="mode", |
| 21 | 243 action="store_const", const=PKK_MODE_XML, |
| 22 | 244 help="output as XML") |
| 0 | 245 |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
246 optparser.add_argument("-n", "--normalize", |
| 22 | 247 dest="normalize", |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
248 action="store_const", const=True, default=False, |
| 22 | 249 help="output NFC normalized Unicode") |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
250 |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
251 optparser.add_argument("-a", "--annotate", |
| 22 | 252 dest="annotate", |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
253 action="store_const", const=True, default=False, |
| 22 | 254 help="annotate strings") |
| 0 | 255 |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
256 optparser.add_argument("-p", "--debug", |
| 22 | 257 dest="debug", action="store_const", const=True, default=False, |
| 258 help=argparse.SUPPRESS) | |
| 259 | |
| 260 optparser.add_argument("-i", "--indent", | |
| 261 dest="indent", | |
| 262 type=int, choices=range(0, 32), default=4, | |
| 263 metavar="n", | |
| 264 help='indent output by <n> characters (default: %(default)s)') | |
| 0 | 265 |
| 266 | |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
267 ### Show help if needed |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
268 pkk_cfg = optparser.parse_args() |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
269 if len(pkk_cfg.filenames) == 0: |
|
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
270 optparser.print_help() |
| 0 | 271 sys.exit(0) |
| 272 | |
| 273 | |
| 6 | 274 ### Handle each input file |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
275 for filename in pkk_cfg.filenames: |
| 0 | 276 # Parse XML file into element tree |
| 277 try: | |
| 278 uxml = xmlET.parse(filename) | |
| 279 except Exception as e: | |
| 280 pkk_fatal(u"SVG/XML parsing failed: {0}".format(str(e))) | |
| 281 | |
| 282 # Dump output | |
| 283 try: | |
| 284 xroot = uxml.getroot() | |
| 285 for dnode in xroot.findall("./DictionaryEntry"): | |
| 7 | 286 |
|
20
f274504eafd0
Use Python argparse module instead of custom self-rolled argument parser.
Matti Hamalainen <ccr@tnsp.org>
parents:
19
diff
changeset
|
287 if pkk_cfg.debug and dnode.attrib["identifier"] not in pkk_debug_list: |
| 7 | 288 continue |
| 289 | |
| 21 | 290 if pkk_cfg.mode == PKK_MODE_NORMAL: |
| 19 | 291 try: |
| 292 pkk_output_node(0, dnode) | |
| 293 except Exception as e: | |
| 294 pkk_dump_recursive(0, dnode) | |
| 295 print(str(e)) | |
| 296 sys.exit(0) | |
| 21 | 297 elif pkk_cfg.mode == PKK_MODE_DUMP: |
| 7 | 298 pkk_dump_recursive(0, dnode) |
| 21 | 299 elif pkk_cfg.mode == PKK_MODE_XML: |
| 10 | 300 pkk_print(str(xmlET.tostring(dnode, encoding="utf8")) + "\n") |
| 0 | 301 else: |
| 10 | 302 pkk_fatal("Invalid operation mode?") |
| 7 | 303 |
| 304 print("\n") | |
| 0 | 305 |
| 306 except (BrokenPipeError, IOError) as e: | |
| 307 sys.stderr.close() | |
| 308 sys.exit(1) | |
| 309 | |
| 310 pkk_cleanup() | |
| 311 sys.exit(0) |