Mercurial > hg > lxmldump
annotate lxmldump.py @ 18:ff959de0f6c8
Add grammatical attributes.
| author | Matti Hamalainen <ccr@tnsp.org> |
|---|---|
| date | Tue, 11 May 2021 15:44:33 +0300 |
| parents | 6fa24c711f86 |
| children | 7c6eb57798bd |
| 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 "annotate": False, | |
| 10 | 26 "mode": 0, |
|
5
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
27 "normalize": False, |
| 7 | 28 "debug": False, |
| 0 | 29 } |
| 30 | |
| 31 | |
| 7 | 32 pkk_str_fmap = { |
| 33 "Fragment" : ["<", ">"], | |
| 34 } | |
| 35 | |
| 36 | |
| 37 pkk_debug_list = [ | |
| 38 "ahas", | |
| 39 "ahavakkaine", | |
| 40 "ahavakala", | |
| 41 "ahavakoittuo", | |
| 42 "ahvaliha", | |
| 9 | 43 "aloilleh", |
| 18 | 44 "hanjahtoakseh", |
| 7 | 45 ] |
| 46 | |
| 47 | |
| 0 | 48 ### |
| 49 ### Misc. helper functions, etc | |
| 50 ### | |
| 51 def pkk_cleanup(): | |
| 52 return 0 | |
| 53 | |
| 54 | |
| 13 | 55 ## 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
|
56 def pkk_print(smsg): |
| 12 | 57 try: |
| 58 if pkk_cfg["normalize"]: | |
| 59 sys.stdout.write(unicodedata.normalize("NFC", smsg)) | |
| 60 else: | |
| 61 sys.stdout.write(smsg) | |
| 62 | |
| 63 except (BrokenPipeError, IOError) as e: | |
| 64 sys.stderr.close() | |
| 65 | |
| 0 | 66 |
| 13 | 67 ## Print string with indentation |
| 7 | 68 def pkk_printi(indent, smsg): |
| 69 pkk_print((" " * indent) + smsg) | |
| 70 | |
| 0 | 71 |
| 13 | 72 ## Check value against current verbosity level |
| 8 | 73 def pkk_verbosity(lvl): |
| 74 return pkk_cfg["verbosity"] >= lvl | |
| 75 | |
| 76 | |
| 0 | 77 ## Fatal error handler |
| 78 def pkk_fatal(smsg): | |
| 79 print(u"ERROR: "+ smsg) | |
| 80 sys.exit(1) | |
| 81 | |
| 82 | |
| 83 ## Handler for SIGINT signals | |
| 84 def pkk_signal_handler(signal, frame): | |
| 85 pkk_cleanup() | |
| 86 print(u"\nQuitting due to SIGINT / Ctrl+C!") | |
| 87 sys.exit(1) | |
| 88 | |
| 89 | |
| 13 | 90 ## Clean string by removing tabs and newlines |
| 10 | 91 def pkk_str_clean(mstr): |
| 92 return re.sub(r'[\n\r\t]', '', mstr) | |
| 93 | |
| 94 | |
| 13 | 95 ## Format "Ptr" node as text |
| 9 | 96 def pkk_ptr_to_text(pnode): |
| 16 | 97 return "<PTR:{}>{}</PTR>".format( |
| 10 | 98 pnode.attrib["{http://www.w3.org/TR/xlink}href"], |
| 99 ("".join(pnode.itertext())).strip()) | |
| 9 | 100 |
| 101 | |
| 13 | 102 ## Get text inside a given node |
| 16 | 103 def pkk_node_to_text(lnode): |
| 7 | 104 stmp = "" |
| 105 for pnode in lnode.iter(): | |
| 9 | 106 if pnode.tag == "Ptr": |
| 107 stmp += pkk_ptr_to_text(pnode) | |
| 108 else: | |
| 109 if isinstance(pnode.text, str): | |
| 10 | 110 ptext = pkk_str_clean(pnode.text).strip() |
| 9 | 111 if pkk_cfg["annotate"] and isinstance(pnode.tag, str) and pnode.tag in pkk_str_fmap: |
| 112 stmp += pkk_str_fmap[pnode.tag][0] + ptext + pkk_str_fmap[pnode.tag][1] | |
| 113 else: | |
| 114 stmp += ptext | |
| 7 | 115 |
| 9 | 116 if isinstance(pnode.tail, str): |
| 10 | 117 stmp += pkk_str_clean(pnode.tail) |
| 7 | 118 |
| 119 return stmp.strip() | |
| 120 | |
| 121 | |
| 13 | 122 ## Simple recursive dump starting at given node |
| 7 | 123 def pkk_dump_recursive(indent, lnode): |
| 124 if lnode.tag in ["Example"]: | |
| 16 | 125 stmp = pkk_node_to_text(lnode) |
| 9 | 126 pkk_printi(indent, "{} \"{}\"\n".format(lnode.tag, stmp)) |
| 6 | 127 else: |
| 7 | 128 if isinstance(lnode.text, str): |
| 10 | 129 stmp = pkk_str_clean(lnode.text).strip() |
| 7 | 130 if stmp != "": |
| 131 stmp = " \""+ stmp +"\"" | |
| 132 else: | |
| 133 stmp = "" | |
| 6 | 134 |
| 135 if len(lnode.attrib) > 0: | |
| 10 | 136 atmp = " "+ str(lnode.attrib) |
| 6 | 137 else: |
| 138 atmp = "" | |
| 139 | |
| 7 | 140 pkk_printi(indent, "{}{}{}\n".format(lnode.tag, atmp, stmp)) |
| 6 | 141 for qnode in lnode.findall("./*"): |
| 7 | 142 pkk_dump_recursive(indent + 1, qnode) |
| 6 | 143 |
| 144 | |
| 17 | 145 ## Output item(s) under given node with given format string |
| 13 | 146 def pkk_output_subs_fmt(indent, dnode, dsub, dname, dfmt): |
| 7 | 147 for qnode in dnode.findall(dsub): |
| 16 | 148 pkk_printi(indent, dfmt.format(dname, pkk_node_to_text(qnode))) |
| 7 | 149 |
| 17 | 150 ## Output item(s) under given node with a prefixed name string |
| 13 | 151 def pkk_output_subs_prefix(indent, dnode, dsub, dname): |
| 152 pkk_output_subs_fmt(indent, dnode, dsub, dname, "{0} \"{1}\"\n") | |
| 7 | 153 |
| 154 | |
| 17 | 155 ## Output a main "Headword" or "Sense" node under it |
| 7 | 156 def pkk_output_sense(indent, dnode): |
| 17 | 157 # Search form and definition |
| 13 | 158 pkk_output_subs_prefix(indent, dnode, "./SearchForm", "srch") |
| 159 pkk_output_subs_prefix(indent, dnode, "./Definition", "defn") | |
| 7 | 160 |
| 17 | 161 # Examples |
| 7 | 162 for wnode in dnode.findall("./ExampleBlock/ExampleCtn"): |
| 16 | 163 sstr = pkk_node_to_text(wnode.find("./Example")) |
| 8 | 164 lstr = "" |
| 6 | 165 |
| 8 | 166 if pkk_verbosity(1): |
| 167 ltmp = [] | |
| 168 for qnode in wnode.findall("./FreeTopic[@type='levikki']/GeographicalUsage"): | |
| 16 | 169 ltmp.append("{} [{}]".format(pkk_node_to_text(qnode), qnode.attrib["class"])) |
| 8 | 170 |
| 171 if len(ltmp) > 0: | |
| 172 lstr = " ({})".format(", ".join(ltmp)) | |
| 7 | 173 |
| 174 pkk_printi(indent + 1, "{} \"{}\"{}\n".format("exmp", sstr, lstr)) | |
| 175 | |
| 176 | |
| 17 | 177 ## Output one "DictionaryEntry" node |
| 7 | 178 def pkk_output_node(indent, dnode): |
| 6 | 179 |
| 7 | 180 for wnode in dnode.findall("./HeadwordCtn"): |
| 18 | 181 # Create list with grammatical attributes (noun, verb, etc.) |
| 182 tmpl = [] | |
| 183 for pnode in wnode.findall("./PartOfSpeechCtn/PartOfSpeech"): | |
| 184 tmpl.append(pnode.attrib["freeValue"]) | |
| 185 | |
| 186 for pnode in wnode.findall("./GrammaticalNote"): | |
| 187 tmpl.append(pnode.text.strip()) | |
| 188 | |
| 189 # Remove duplicates and sort the list | |
| 190 tmpl = list(set(tmpl)) | |
| 191 tmpl.sort(reverse=False, key=lambda attr: (attr, len(attr))) | |
| 192 | |
| 193 # Print the headword and attributes if any | |
| 194 pkk_output_subs_fmt(indent, wnode, "./Headword", "", "\"{1}\"") | |
| 195 if len(tmpl) > 0: | |
| 196 pkk_print(" ({})\n".format(" ; ".join(tmpl))) | |
| 197 else: | |
| 198 pkk_print("\n") | |
| 199 | |
| 200 # Print main "sense" | |
| 7 | 201 pkk_output_sense(indent + 1, wnode) |
| 6 | 202 |
| 18 | 203 # Print any other "senses" |
| 7 | 204 index = 1 |
| 205 for wnode in dnode.findall("./SenseGrp"): | |
| 8 | 206 pkk_printi(indent + 1, "sense #{}\n".format(index)) |
| 7 | 207 pkk_output_sense(indent + 2, wnode) |
| 208 index += 1 | |
| 6 | 209 |
| 210 | |
| 0 | 211 ### |
| 212 ### Main program starts | |
| 213 ### | |
| 214 signal.signal(signal.SIGINT, pkk_signal_handler) | |
| 215 | |
| 216 | |
| 217 ### Check if we have arguments | |
| 218 pkk_show_help = False | |
| 219 pkk_filenames = [] | |
| 220 argc = 1 | |
| 221 while argc < len(sys.argv): | |
| 222 arg = sys.argv[argc] | |
| 223 | |
| 224 needs_param = False | |
| 225 if argc + 1 < len(sys.argv): | |
| 226 param = sys.argv[argc + 1] | |
| 227 else: | |
| 228 param = None | |
| 229 | |
| 230 # Check for option type arg | |
| 231 if arg[0:1] == "-": | |
| 232 oarg = arg | |
| 233 arg = arg.lstrip("-") | |
| 234 | |
| 235 if arg == "help" or arg == "h": | |
| 236 pkk_show_help = True | |
| 237 elif arg == "dump" or arg == "d": | |
| 10 | 238 pkk_cfg["mode"] = 1 |
| 239 elif arg == "xml" or arg == "x": | |
| 240 pkk_cfg["mode"] = 2 | |
|
5
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
241 elif arg == "normalize" or arg == "n": |
|
274b2091137c
Some more work on cleaning this up.
Matti Hamalainen <ccr@tnsp.org>
parents:
4
diff
changeset
|
242 pkk_cfg["normalize"] = True |
| 8 | 243 elif arg == "annotate" or arg == "a": |
| 244 pkk_cfg["annotate"] = True | |
| 7 | 245 elif arg == "p": |
| 246 pkk_cfg["debug"] = True | |
| 8 | 247 elif arg == "verbosity" or arg == "v": |
| 248 needs_param = True | |
| 249 pkk_cfg["verbosity"] = param | |
| 0 | 250 else: |
| 251 pkk_fatal(u"Invalid option argument '{0}'.".format(oarg)) | |
| 252 | |
| 253 if needs_param and param == None: | |
| 254 pkk_fatal(u"Option '{0}' requires an argument.".format(oarg)) | |
| 255 else: | |
| 256 # Non-option argument | |
| 257 pkk_filenames.append(arg) | |
| 258 | |
| 259 if needs_param: | |
| 260 argc += 2 | |
| 261 else: | |
| 262 argc += 1 | |
| 263 | |
| 264 | |
| 265 ### Show help if requested | |
| 266 if pkk_show_help or len(pkk_filenames) == 0: | |
| 267 print(u"lxmldump - Dump ISO/FDIS 1951 XML file data") | |
| 268 print(u"Usage: {0} <options> <input xml file(s)>". | |
| 269 format(str(Path(sys.argv[0]).name))) | |
| 270 print(u"") | |
| 271 print(u" --help Show this help") | |
| 15 | 272 print(u" -d, --dump Output as simple dump") |
| 273 print(u" -x, --xml Output as XML") | |
| 6 | 274 print(u" -n, --normalize Output NFC normalized Unicode") |
| 8 | 275 print(u" -a, --annotate Annotate strings") |
| 276 print(u" -v, --verbosity <n> Set verbosity level (0 - 3)") | |
| 0 | 277 print(u"") |
| 278 sys.exit(0) | |
| 279 | |
| 280 | |
| 8 | 281 ### Validate settings |
| 282 try: | |
| 283 pkk_cfg["verbosity"] = int(pkk_cfg["verbosity"]) | |
| 284 except Exception as e: | |
| 285 pkk_fatal(u"Verbosity level is not a valid integer.") | |
| 286 if pkk_cfg["verbosity"] < 0 or pkk_cfg["verbosity"] > 3: | |
| 287 pkk_fatal(u"Invalid verbosity level value {0}.".format(pkk_cfg["verbosity"])) | |
| 288 | |
| 289 | |
| 6 | 290 ### Handle each input file |
| 0 | 291 for filename in pkk_filenames: |
| 292 # Parse XML file into element tree | |
| 293 try: | |
| 294 uxml = xmlET.parse(filename) | |
| 295 except Exception as e: | |
| 296 pkk_fatal(u"SVG/XML parsing failed: {0}".format(str(e))) | |
| 297 | |
| 298 # Dump output | |
| 299 try: | |
| 300 xroot = uxml.getroot() | |
| 301 for dnode in xroot.findall("./DictionaryEntry"): | |
| 7 | 302 |
| 303 if pkk_cfg["debug"] and dnode.attrib["identifier"] not in pkk_debug_list: | |
| 304 continue | |
| 305 | |
| 10 | 306 if pkk_cfg["mode"] == 0: |
| 307 pkk_output_node(0, dnode) | |
| 308 elif pkk_cfg["mode"] == 1: | |
| 7 | 309 pkk_dump_recursive(0, dnode) |
| 10 | 310 elif pkk_cfg["mode"] == 2: |
| 311 pkk_print(str(xmlET.tostring(dnode, encoding="utf8")) + "\n") | |
| 0 | 312 else: |
| 10 | 313 pkk_fatal("Invalid operation mode?") |
| 7 | 314 |
| 315 print("\n") | |
| 0 | 316 |
| 317 except (BrokenPipeError, IOError) as e: | |
| 318 sys.stderr.close() | |
| 319 sys.exit(1) | |
| 320 | |
| 321 pkk_cleanup() | |
| 322 sys.exit(0) |