|
| 1 | +"""" Command Line Interface for `openml` to configure its settings. """ |
| 2 | + |
| 3 | +import argparse |
| 4 | +import os |
| 5 | +import pathlib |
| 6 | +import string |
| 7 | +from typing import Union, Callable |
| 8 | +from urllib.parse import urlparse |
| 9 | + |
| 10 | + |
| 11 | +from openml import config |
| 12 | + |
| 13 | + |
| 14 | +def is_hex(string_: str) -> bool: |
| 15 | + return all(c in string.hexdigits for c in string_) |
| 16 | + |
| 17 | + |
| 18 | +def looks_like_url(url: str) -> bool: |
| 19 | + # There's no thorough url parser, but we only seem to use netloc. |
| 20 | + try: |
| 21 | + return bool(urlparse(url).netloc) |
| 22 | + except Exception: |
| 23 | + return False |
| 24 | + |
| 25 | + |
| 26 | +def wait_until_valid_input( |
| 27 | + prompt: str, check: Callable[[str], str], sanitize: Union[Callable[[str], str], None] |
| 28 | +) -> str: |
| 29 | + """ Asks `prompt` until an input is received which returns True for `check`. |
| 30 | +
|
| 31 | + Parameters |
| 32 | + ---------- |
| 33 | + prompt: str |
| 34 | + message to display |
| 35 | + check: Callable[[str], str] |
| 36 | + function to call with the given input, that provides an error message if the input is not |
| 37 | + valid otherwise, and False-like otherwise. |
| 38 | + sanitize: Callable[[str], str], optional |
| 39 | + A function which attempts to sanitize the user input (e.g. auto-complete). |
| 40 | +
|
| 41 | + Returns |
| 42 | + ------- |
| 43 | + valid input |
| 44 | +
|
| 45 | + """ |
| 46 | + |
| 47 | + while True: |
| 48 | + response = input(prompt) |
| 49 | + if sanitize: |
| 50 | + response = sanitize(response) |
| 51 | + error_message = check(response) |
| 52 | + if error_message: |
| 53 | + print(error_message, end="\n\n") |
| 54 | + else: |
| 55 | + return response |
| 56 | + |
| 57 | + |
| 58 | +def print_configuration(): |
| 59 | + file = config.determine_config_file_path() |
| 60 | + header = f"File '{file}' contains (or defaults to):" |
| 61 | + print(header) |
| 62 | + |
| 63 | + max_key_length = max(map(len, config.get_config_as_dict())) |
| 64 | + for field, value in config.get_config_as_dict().items(): |
| 65 | + print(f"{field.ljust(max_key_length)}: {value}") |
| 66 | + |
| 67 | + |
| 68 | +def verbose_set(field, value): |
| 69 | + config.set_field_in_config_file(field, value) |
| 70 | + print(f"{field} set to '{value}'.") |
| 71 | + |
| 72 | + |
| 73 | +def configure_apikey(value: str) -> None: |
| 74 | + def check_apikey(apikey: str) -> str: |
| 75 | + if len(apikey) != 32: |
| 76 | + return f"The key should contain 32 characters but contains {len(apikey)}." |
| 77 | + if not is_hex(apikey): |
| 78 | + return "Some characters are not hexadecimal." |
| 79 | + return "" |
| 80 | + |
| 81 | + instructions = ( |
| 82 | + f"Your current API key is set to: '{config.apikey}'. " |
| 83 | + "You can get an API key at https://new.openml.org. " |
| 84 | + "You must create an account if you don't have one yet:\n" |
| 85 | + " 1. Log in with the account.\n" |
| 86 | + " 2. Navigate to the profile page (top right circle > Your Profile). \n" |
| 87 | + " 3. Click the API Key button to reach the page with your API key.\n" |
| 88 | + "If you have any difficulty following these instructions, let us know on Github." |
| 89 | + ) |
| 90 | + |
| 91 | + configure_field( |
| 92 | + field="apikey", |
| 93 | + value=value, |
| 94 | + check_with_message=check_apikey, |
| 95 | + intro_message=instructions, |
| 96 | + input_message="Please enter your API key:", |
| 97 | + ) |
| 98 | + |
| 99 | + |
| 100 | +def configure_server(value: str) -> None: |
| 101 | + def check_server(server: str) -> str: |
| 102 | + is_shorthand = server in ["test", "production"] |
| 103 | + if is_shorthand or looks_like_url(server): |
| 104 | + return "" |
| 105 | + return "Must be 'test', 'production' or a url." |
| 106 | + |
| 107 | + def replace_shorthand(server: str) -> str: |
| 108 | + if server == "test": |
| 109 | + return "https://test.openml.org/api/v1/xml" |
| 110 | + if server == "production": |
| 111 | + return "https://www.openml.org/api/v1/xml" |
| 112 | + return server |
| 113 | + |
| 114 | + configure_field( |
| 115 | + field="server", |
| 116 | + value=value, |
| 117 | + check_with_message=check_server, |
| 118 | + intro_message="Specify which server you wish to connect to.", |
| 119 | + input_message="Specify a url or use 'test' or 'production' as a shorthand: ", |
| 120 | + sanitize=replace_shorthand, |
| 121 | + ) |
| 122 | + |
| 123 | + |
| 124 | +def configure_cachedir(value: str) -> None: |
| 125 | + def check_cache_dir(path: str) -> str: |
| 126 | + p = pathlib.Path(path) |
| 127 | + if p.is_file(): |
| 128 | + return f"'{path}' is a file, not a directory." |
| 129 | + expanded = p.expanduser() |
| 130 | + if not expanded.is_absolute(): |
| 131 | + return f"'{path}' is not absolute (even after expanding '~')." |
| 132 | + if not expanded.exists(): |
| 133 | + try: |
| 134 | + os.mkdir(expanded) |
| 135 | + except PermissionError: |
| 136 | + return f"'{path}' does not exist and there are not enough permissions to create it." |
| 137 | + return "" |
| 138 | + |
| 139 | + configure_field( |
| 140 | + field="cachedir", |
| 141 | + value=value, |
| 142 | + check_with_message=check_cache_dir, |
| 143 | + intro_message="Configuring the cache directory. It can not be a relative path.", |
| 144 | + input_message="Specify the directory to use (or create) as cache directory: ", |
| 145 | + ) |
| 146 | + print("NOTE: Data from your old cache directory is not moved over.") |
| 147 | + |
| 148 | + |
| 149 | +def configure_connection_n_retries(value: str) -> None: |
| 150 | + def valid_connection_retries(n: str) -> str: |
| 151 | + if not n.isdigit(): |
| 152 | + return f"Must be an integer number (smaller than {config.max_retries})." |
| 153 | + if int(n) > config.max_retries: |
| 154 | + return f"connection_n_retries may not exceed {config.max_retries}." |
| 155 | + if int(n) == 0: |
| 156 | + return "connection_n_retries must be non-zero." |
| 157 | + return "" |
| 158 | + |
| 159 | + configure_field( |
| 160 | + field="connection_n_retries", |
| 161 | + value=value, |
| 162 | + check_with_message=valid_connection_retries, |
| 163 | + intro_message="Configuring the number of times to attempt to connect to the OpenML Server", |
| 164 | + input_message=f"Enter an integer between 0 and {config.max_retries}: ", |
| 165 | + ) |
| 166 | + |
| 167 | + |
| 168 | +def configure_avoid_duplicate_runs(value: str) -> None: |
| 169 | + def is_python_bool(bool_: str) -> str: |
| 170 | + if bool_ in ["True", "False"]: |
| 171 | + return "" |
| 172 | + return "Must be 'True' or 'False' (mind the capital)." |
| 173 | + |
| 174 | + def autocomplete_bool(bool_: str) -> str: |
| 175 | + if bool_.lower() in ["n", "no", "f", "false", "0"]: |
| 176 | + return "False" |
| 177 | + if bool_.lower() in ["y", "yes", "t", "true", "1"]: |
| 178 | + return "True" |
| 179 | + return bool_ |
| 180 | + |
| 181 | + intro_message = ( |
| 182 | + "If set to True, when `run_flow_on_task` or similar methods are called a lookup is " |
| 183 | + "performed to see if there already exists such a run on the server. " |
| 184 | + "If so, download those results instead. " |
| 185 | + "If set to False, runs will always be executed." |
| 186 | + ) |
| 187 | + |
| 188 | + configure_field( |
| 189 | + field="avoid_duplicate_runs", |
| 190 | + value=value, |
| 191 | + check_with_message=is_python_bool, |
| 192 | + intro_message=intro_message, |
| 193 | + input_message="Enter 'True' or 'False': ", |
| 194 | + sanitize=autocomplete_bool, |
| 195 | + ) |
| 196 | + |
| 197 | + |
| 198 | +def configure_verbosity(value: str) -> None: |
| 199 | + def is_zero_through_two(verbosity: str) -> str: |
| 200 | + if verbosity in ["0", "1", "2"]: |
| 201 | + return "" |
| 202 | + return "Must be '0', '1' or '2'." |
| 203 | + |
| 204 | + intro_message = ( |
| 205 | + "Set the verbosity of log messages which should be shown by openml-python." |
| 206 | + " 0: normal output (warnings and errors)" |
| 207 | + " 1: info output (some high-level progress output)" |
| 208 | + " 2: debug output (detailed information (for developers))" |
| 209 | + ) |
| 210 | + |
| 211 | + configure_field( |
| 212 | + field="verbosity", |
| 213 | + value=value, |
| 214 | + check_with_message=is_zero_through_two, |
| 215 | + intro_message=intro_message, |
| 216 | + input_message="Enter '0', '1' or '2': ", |
| 217 | + ) |
| 218 | + |
| 219 | + |
| 220 | +def configure_field( |
| 221 | + field: str, |
| 222 | + value: Union[None, str], |
| 223 | + check_with_message: Callable[[str], str], |
| 224 | + intro_message: str, |
| 225 | + input_message: str, |
| 226 | + sanitize: Union[Callable[[str], str], None] = None, |
| 227 | +) -> None: |
| 228 | + """ Configure `field` with `value`. If `value` is None ask the user for input. |
| 229 | +
|
| 230 | + `value` and user input are first corrected/auto-completed with `convert_value` if provided, |
| 231 | + then validated with `check_with_message` function. |
| 232 | + If the user input a wrong value in interactive mode, the user gets to input a new value. |
| 233 | + The new valid value is saved in the openml configuration file. |
| 234 | + In case an invalid `value` is supplied directly (non-interactive), no changes are made. |
| 235 | +
|
| 236 | + Parameters |
| 237 | + ---------- |
| 238 | + field: str |
| 239 | + Field to set. |
| 240 | + value: str, None |
| 241 | + Value to field to. If `None` will ask user for input. |
| 242 | + check_with_message: Callable[[str], str] |
| 243 | + Function which validates `value` or user input, and returns either an error message if it |
| 244 | + is invalid, or a False-like value if `value` is valid. |
| 245 | + intro_message: str |
| 246 | + Message that is printed once if user input is requested (e.g. instructions). |
| 247 | + input_message: str |
| 248 | + Message that comes with the input prompt. |
| 249 | + sanitize: Union[Callable[[str], str], None] |
| 250 | + A function to convert user input to 'more acceptable' input, e.g. for auto-complete. |
| 251 | + If no correction of user input is possible, return the original value. |
| 252 | + If no function is provided, don't attempt to correct/auto-complete input. |
| 253 | + """ |
| 254 | + if value is not None: |
| 255 | + if sanitize: |
| 256 | + value = sanitize(value) |
| 257 | + malformed_input = check_with_message(value) |
| 258 | + if malformed_input: |
| 259 | + print(malformed_input) |
| 260 | + quit() |
| 261 | + else: |
| 262 | + print(intro_message) |
| 263 | + value = wait_until_valid_input( |
| 264 | + prompt=input_message, check=check_with_message, sanitize=sanitize, |
| 265 | + ) |
| 266 | + verbose_set(field, value) |
| 267 | + |
| 268 | + |
| 269 | +def configure(args: argparse.Namespace): |
| 270 | + """ Calls the right submenu(s) to edit `args.field` in the configuration file. """ |
| 271 | + set_functions = { |
| 272 | + "apikey": configure_apikey, |
| 273 | + "server": configure_server, |
| 274 | + "cachedir": configure_cachedir, |
| 275 | + "connection_n_retries": configure_connection_n_retries, |
| 276 | + "avoid_duplicate_runs": configure_avoid_duplicate_runs, |
| 277 | + "verbosity": configure_verbosity, |
| 278 | + } |
| 279 | + |
| 280 | + def not_supported_yet(_): |
| 281 | + print(f"Setting '{args.field}' is not supported yet.") |
| 282 | + |
| 283 | + if args.field not in ["all", "none"]: |
| 284 | + set_functions.get(args.field, not_supported_yet)(args.value) |
| 285 | + else: |
| 286 | + if args.value is not None: |
| 287 | + print(f"Can not set value ('{args.value}') when field is specified as '{args.field}'.") |
| 288 | + quit() |
| 289 | + print_configuration() |
| 290 | + |
| 291 | + if args.field == "all": |
| 292 | + for set_field_function in set_functions.values(): |
| 293 | + print() # Visually separating the output by field. |
| 294 | + set_field_function(args.value) |
| 295 | + |
| 296 | + |
| 297 | +def main() -> None: |
| 298 | + subroutines = {"configure": configure} |
| 299 | + |
| 300 | + parser = argparse.ArgumentParser() |
| 301 | + subparsers = parser.add_subparsers(dest="subroutine") |
| 302 | + |
| 303 | + parser_configure = subparsers.add_parser( |
| 304 | + "configure", |
| 305 | + description="Set or read variables in your configuration file. For more help also see " |
| 306 | + "'https://openml.github.io/openml-python/master/usage.html#configuration'.", |
| 307 | + ) |
| 308 | + |
| 309 | + configurable_fields = [f for f in config._defaults if f not in ["max_retries"]] |
| 310 | + |
| 311 | + parser_configure.add_argument( |
| 312 | + "field", |
| 313 | + type=str, |
| 314 | + choices=[*configurable_fields, "all", "none"], |
| 315 | + default="all", |
| 316 | + nargs="?", |
| 317 | + help="The field you wish to edit. " |
| 318 | + "Choosing 'all' lets you configure all fields one by one. " |
| 319 | + "Choosing 'none' will print out the current configuration.", |
| 320 | + ) |
| 321 | + |
| 322 | + parser_configure.add_argument( |
| 323 | + "value", type=str, default=None, nargs="?", help="The value to set the FIELD to.", |
| 324 | + ) |
| 325 | + |
| 326 | + args = parser.parse_args() |
| 327 | + subroutines.get(args.subroutine, lambda _: parser.print_help())(args) |
| 328 | + |
| 329 | + |
| 330 | +if __name__ == "__main__": |
| 331 | + main() |
0 commit comments