Mercurial > hg > lxmldump
annotate lxmldump.py @ 9:2a8c65d22f86
Moar.
| author | Matti Hamalainen <ccr@tnsp.org> |
|---|---|
| date | Tue, 11 May 2021 09:33:21 +0300 |
| parents | ce07bb2a247b |
| children | 013f0cd9e5b3 |
| 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 ### | |
| 8 ### Python 3.7+ required! | |
| 0 | 9 ### |
| 10 import sys | |
| 11 import signal | |
| 12 import re | |
| 13 from pathlib import Path | |
| 14 import xml.etree.ElementTree as xmlET | |
|
5
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
15 import unicodedata |
| 0 | 16 |
| 17 assert sys.version_info >= (3, 7) | |
| 18 | |
| 19 | |
| 20 ### | |
| 21 ### Default settings | |
| 22 ### | |
| 23 pkk_cfg = { | |
| 8 | 24 "verbosity": 3, |
| 25 | |
| 26 "annotate": False, | |
| 0 | 27 "dump": False, |
|
5
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
28 "normalize": False, |
| 7 | 29 |
| 30 "debug": False, | |
| 0 | 31 } |
| 32 | |
| 33 | |
| 7 | 34 pkk_str_fmap = { |
| 35 "Fragment" : ["<", ">"], | |
| 36 } | |
| 37 | |
| 38 | |
| 39 pkk_debug_list = [ | |
| 40 "ahas", | |
| 41 "ahavakkaine", | |
| 42 "ahavakala", | |
| 43 "ahavakoittuo", | |
| 44 "ahvaliha", | |
| 9 | 45 "aloilleh", |
| 7 | 46 ] |
| 47 | |
| 48 | |
| 0 | 49 ### |
| 50 ### Misc. helper functions, etc | |
| 51 ### | |
| 52 def pkk_cleanup(): | |
| 53 return 0 | |
| 54 | |
| 55 | |
| 56 ## Wrapper for print() | |
|
5
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
57 def pkk_print(smsg): |
|
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
58 if pkk_cfg["normalize"]: |
|
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
59 sys.stdout.write(unicodedata.normalize("NFC", smsg)) |
|
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
60 else: |
|
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
61 sys.stdout.write(smsg) |
| 0 | 62 |
| 7 | 63 def pkk_printi(indent, smsg): |
| 64 pkk_print((" " * indent) + smsg) | |
| 65 | |
| 0 | 66 |
| 8 | 67 def pkk_verbosity(lvl): |
| 68 return pkk_cfg["verbosity"] >= lvl | |
| 69 | |
| 70 | |
| 0 | 71 ## Fatal error handler |
| 72 def pkk_fatal(smsg): | |
| 73 print(u"ERROR: "+ smsg) | |
| 74 sys.exit(1) | |
| 75 | |
| 76 | |
| 77 ## Handler for SIGINT signals | |
| 78 def pkk_signal_handler(signal, frame): | |
| 79 pkk_cleanup() | |
| 80 print(u"\nQuitting due to SIGINT / Ctrl+C!") | |
| 81 sys.exit(1) | |
| 82 | |
| 83 | |
| 9 | 84 ## |
| 85 def pkk_ptr_to_text(pnode): | |
| 86 # return "PTR: <{}>".format("".join(pnode.itertext()).strip()) | |
| 87 return "PTR: <{}>".format(xmlET.tostring(pnode)) | |
| 88 | |
| 89 | |
| 7 | 90 def pkk_get_text(lnode): |
| 91 stmp = "" | |
| 92 for pnode in lnode.iter(): | |
| 9 | 93 if pnode.tag == "Ptr": |
| 94 stmp += pkk_ptr_to_text(pnode) | |
| 95 else: | |
| 96 if isinstance(pnode.text, str): | |
| 97 ptext = pnode.text | |
| 98 if pkk_cfg["annotate"] and isinstance(pnode.tag, str) and pnode.tag in pkk_str_fmap: | |
| 99 stmp += pkk_str_fmap[pnode.tag][0] + ptext + pkk_str_fmap[pnode.tag][1] | |
| 100 else: | |
| 101 stmp += ptext | |
| 7 | 102 |
| 9 | 103 if isinstance(pnode.tail, str): |
| 104 stmp += " "+ pnode.tail.strip() | |
| 7 | 105 |
| 106 return stmp.strip() | |
| 107 | |
| 108 | |
| 6 | 109 ## |
| 7 | 110 def pkk_dump_recursive(indent, lnode): |
| 111 if lnode.tag in ["Example"]: | |
| 112 stmp = pkk_get_text(lnode) | |
| 9 | 113 pkk_printi(indent, "{} \"{}\"\n".format(lnode.tag, stmp)) |
| 6 | 114 else: |
| 7 | 115 if isinstance(lnode.text, str): |
| 116 stmp = lnode.text.strip() | |
| 117 if stmp != "": | |
| 118 stmp = " \""+ stmp +"\"" | |
| 119 else: | |
| 120 stmp = "" | |
| 6 | 121 |
| 122 if len(lnode.attrib) > 0: | |
| 123 atmp = " "+str(lnode.attrib) | |
| 124 else: | |
| 125 atmp = "" | |
| 126 | |
| 7 | 127 pkk_printi(indent, "{}{}{}\n".format(lnode.tag, atmp, stmp)) |
| 6 | 128 for qnode in lnode.findall("./*"): |
| 7 | 129 pkk_dump_recursive(indent + 1, qnode) |
| 6 | 130 |
| 131 | |
| 132 ## | |
| 7 | 133 def pkk_output_one(indent, dnode, dsub): |
| 134 for qnode in dnode.findall(dsub): | |
| 135 pkk_printi(indent, "{}\n".format(pkk_get_text(qnode))) | |
| 136 | |
| 137 def pkk_output_subs(indent, dnode, dsub, dname): | |
| 138 for qnode in dnode.findall(dsub): | |
| 139 pkk_printi(indent, "{} \"{}\"\n".format(dname, pkk_get_text(qnode))) | |
| 140 | |
| 141 | |
| 142 def pkk_output_sense(indent, dnode): | |
| 143 pkk_output_subs(indent, dnode, "./SearchForm", "srch") | |
| 144 pkk_output_subs(indent, dnode, "./Definition", "defn") | |
| 145 | |
| 146 for wnode in dnode.findall("./ExampleBlock/ExampleCtn"): | |
| 147 sstr = pkk_get_text(wnode.find("./Example")) | |
| 8 | 148 lstr = "" |
| 6 | 149 |
| 8 | 150 if pkk_verbosity(1): |
| 151 ltmp = [] | |
| 152 for qnode in wnode.findall("./FreeTopic[@type='levikki']/GeographicalUsage"): | |
| 153 ltmp.append("{} [{}]".format(pkk_get_text(qnode), qnode.attrib["class"])) | |
| 154 | |
| 155 if len(ltmp) > 0: | |
| 156 lstr = " ({})".format(", ".join(ltmp)) | |
| 7 | 157 |
| 158 pkk_printi(indent + 1, "{} \"{}\"{}\n".format("exmp", sstr, lstr)) | |
| 159 | |
| 160 | |
| 161 def pkk_output_node(indent, dnode): | |
| 6 | 162 |
| 7 | 163 for wnode in dnode.findall("./HeadwordCtn"): |
| 164 pkk_output_one (indent, wnode, "./Headword") | |
| 165 pkk_output_sense(indent + 1, wnode) | |
| 6 | 166 |
| 7 | 167 index = 1 |
| 168 for wnode in dnode.findall("./SenseGrp"): | |
| 8 | 169 pkk_printi(indent + 1, "sense #{}\n".format(index)) |
| 7 | 170 pkk_output_sense(indent + 2, wnode) |
| 171 index += 1 | |
| 6 | 172 |
| 173 | |
| 0 | 174 ### |
| 175 ### Main program starts | |
| 176 ### | |
| 177 signal.signal(signal.SIGINT, pkk_signal_handler) | |
| 178 | |
| 179 | |
| 180 ### Check if we have arguments | |
| 181 pkk_show_help = False | |
| 182 pkk_filenames = [] | |
| 183 argc = 1 | |
| 184 while argc < len(sys.argv): | |
| 185 arg = sys.argv[argc] | |
| 186 | |
| 187 needs_param = False | |
| 188 if argc + 1 < len(sys.argv): | |
| 189 param = sys.argv[argc + 1] | |
| 190 else: | |
| 191 param = None | |
| 192 | |
| 193 # Check for option type arg | |
| 194 if arg[0:1] == "-": | |
| 195 oarg = arg | |
| 196 arg = arg.lstrip("-") | |
| 197 | |
| 198 if arg == "help" or arg == "h": | |
| 199 pkk_show_help = True | |
| 200 elif arg == "dump" or arg == "d": | |
| 201 pkk_cfg["dump"] = True | |
|
5
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
202 elif arg == "normalize" or arg == "n": |
|
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
203 pkk_cfg["normalize"] = True |
| 8 | 204 elif arg == "annotate" or arg == "a": |
| 205 pkk_cfg["annotate"] = True | |
| 7 | 206 elif arg == "p": |
| 207 pkk_cfg["debug"] = True | |
| 8 | 208 elif arg == "verbosity" or arg == "v": |
| 209 needs_param = True | |
| 210 pkk_cfg["verbosity"] = param | |
| 0 | 211 else: |
| 212 pkk_fatal(u"Invalid option argument '{0}'.".format(oarg)) | |
| 213 | |
| 214 if needs_param and param == None: | |
| 215 pkk_fatal(u"Option '{0}' requires an argument.".format(oarg)) | |
| 216 else: | |
| 217 # Non-option argument | |
| 218 pkk_filenames.append(arg) | |
| 219 | |
| 220 if needs_param: | |
| 221 argc += 2 | |
| 222 else: | |
| 223 argc += 1 | |
| 224 | |
| 225 | |
| 226 ### Show help if requested | |
| 227 if pkk_show_help or len(pkk_filenames) == 0: | |
| 228 print(u"lxmldump - Dump ISO/FDIS 1951 XML file data") | |
| 229 print(u"Usage: {0} <options> <input xml file(s)>". | |
| 230 format(str(Path(sys.argv[0]).name))) | |
| 231 print(u"") | |
| 232 print(u" --help Show this help") | |
| 233 print(u" -d, --dump Dump mode") | |
| 6 | 234 print(u" -n, --normalize Output NFC normalized Unicode") |
| 8 | 235 print(u" -a, --annotate Annotate strings") |
| 236 print(u" -v, --verbosity <n> Set verbosity level (0 - 3)") | |
| 0 | 237 print(u"") |
| 238 sys.exit(0) | |
| 239 | |
| 240 | |
| 8 | 241 ### Validate settings |
| 242 try: | |
| 243 pkk_cfg["verbosity"] = int(pkk_cfg["verbosity"]) | |
| 244 except Exception as e: | |
| 245 pkk_fatal(u"Verbosity level is not a valid integer.") | |
| 246 if pkk_cfg["verbosity"] < 0 or pkk_cfg["verbosity"] > 3: | |
| 247 pkk_fatal(u"Invalid verbosity level value {0}.".format(pkk_cfg["verbosity"])) | |
| 248 | |
| 249 | |
| 6 | 250 ### Handle each input file |
| 0 | 251 for filename in pkk_filenames: |
| 252 # Parse XML file into element tree | |
| 253 try: | |
| 254 uxml = xmlET.parse(filename) | |
| 255 except Exception as e: | |
| 256 pkk_fatal(u"SVG/XML parsing failed: {0}".format(str(e))) | |
| 257 | |
| 258 # Dump output | |
| 259 try: | |
| 260 xroot = uxml.getroot() | |
| 261 for dnode in xroot.findall("./DictionaryEntry"): | |
| 7 | 262 |
| 263 if pkk_cfg["debug"] and dnode.attrib["identifier"] not in pkk_debug_list: | |
| 264 continue | |
| 265 | |
| 0 | 266 if pkk_cfg["dump"]: |
| 7 | 267 pkk_dump_recursive(0, dnode) |
| 0 | 268 else: |
| 7 | 269 pkk_output_node(0, dnode) |
| 270 | |
| 271 print("\n") | |
| 0 | 272 |
| 273 except (BrokenPipeError, IOError) as e: | |
| 274 sys.stderr.close() | |
| 275 sys.exit(1) | |
| 276 | |
| 277 pkk_cleanup() | |
| 278 sys.exit(0) |