python_packaging/src/gmxapi/commandline.py

   1 #
   2 # This file is part of the GROMACS molecular simulation package.
   3 #
   4 # Copyright (c) 2019,2020, by the GROMACS development team, led by
   5 # Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6 # and including many others, as listed in the AUTHORS file in the
   7 # top-level source directory and at http://www.gromacs.org.
   8 #
   9 # GROMACS is free software; you can redistribute it and/or
  10 # modify it under the terms of the GNU Lesser General Public License
  11 # as published by the Free Software Foundation; either version 2.1
  12 # of the License, or (at your option) any later version.
  13 #
  14 # GROMACS is distributed in the hope that it will be useful,
  15 # but WITHOUT ANY WARRANTY; without even the implied warranty of
  16 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17 # Lesser General Public License for more details.
  18 #
  19 # You should have received a copy of the GNU Lesser General Public
  20 # License along with GROMACS; if not, see
  21 # http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22 # Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23 #
  24 # If you want to redistribute modifications to GROMACS, please
  25 # consider that scientific software is very special. Version
  26 # control is crucial - bugs must be traceable. We will be happy to
  27 # consider code for inclusion in the official distribution, but
  28 # derived work must not be called official GROMACS. Details are found
  29 # in the README & COPYING files - if they are missing, get the
  30 # official version at http://www.gromacs.org.
  31 #
  32 # To help us fund GROMACS development, we humbly ask that you cite
  33 # the research papers on the package. Check out http://www.gromacs.org.
  34
  35 """
  36 Provide command line operation.
  37 """
  38
  39 __all__ = ['commandline_operation']
  40
  41 import os
  42 import shutil
  43 import subprocess
  44
  45 import gmxapi as gmx
  46 from gmxapi import exceptions
  47 from gmxapi import logger as root_logger
  48 from gmxapi.datamodel import NDArray
  49 from gmxapi.operation import OutputCollectionDescription
  50
  51 # Module-level logger
  52 logger = root_logger.getChild('commandline')
  53 logger.info('Importing {}'.format(__name__))
  54
  55
  56 # Create an Operation that consumes a list and a boolean to produce a string and an integer.
  57 #
  58 # Wrap the defined function using a decorator that
  59 #    * strips the `output` parameter from the signature
  60 #    * provides `output` publishing proxy to the inner function and
  61 #    * produce a result with attributes for
  62 #       * file: mapping of output flags to output filenames
  63 #       * erroroutput: text results in case of error
  64 #       * returncode: integer return code of wrapped command
  65 #
  66 # Note that the existence of the 'file' output map is expressed here, but
  67 # the keys of the map are not implicit or set by the wrapped function.
  68 # For the map to be non-empty, it must be defined before the resulting helper
  69 # function is called.
  70 #
  71 # TODO: Operation returns the output object when called with the shorter signature.
  72 #
  73 @gmx.function_wrapper(output={'erroroutput': str, 'returncode': int})
  74 def cli(command: NDArray, shell: bool, output: OutputCollectionDescription, stdin: str = ''):
  75     """Execute a command line program in a subprocess.
  76
  77     Configure an executable in a subprocess. Executes when run in an execution
  78     Context, as part of a work graph or via gmx.run(). Runs in the current
  79     working directory.
  80
  81     Shell processing is not enabled, but can be considered for a future version.
  82     This means that shell expansions such as environment variables, globbing (`*`),
  83     and other special symbols (like `~` for home directory) are not available.
  84     This allows a simpler and more robust implementation, as well as a better
  85     ability to uniquely identify the effects of a command line operation. If you
  86     think this disallows important use cases, please let us know.
  87
  88     Arguments:
  89          command: a tuple (or list) to be the subprocess arguments, including `executable`
  90          output: mapping of command line flags to output filename arguments
  91          shell: unused (provides forward-compatibility)
  92          stdin (str): String input to send to STDIN (terminal input) of the executable.
  93
  94     Multi-line text sent to *stdin* should be joined into a single string
  95     (e.g. ``'\n'.join(list_of_strings) + '\n'``).
  96     If multiple strings are provided to *stdin*, gmxapi will assume an ensemble,
  97     and will run one operation for each provided string.
  98
  99     Only string input (:py:func:str) to *stdin* is currently supported.
 100     If you have a use case that requires streaming input or binary input,
 101     please open an issue or contact the author(s).
 102
 103     Arguments are iteratively added to the command line with standard Python
 104     iteration, so you should use a tuple or list even if you have only one parameter.
 105     I.e. If you provide a string with `arguments="asdf"` then it will be passed as
 106     `... "a" "s" "d" "f"`. To pass a single string argument, `arguments=("asdf")`
 107     or `arguments=["asdf"]`.
 108
 109     `input` and `output` should be a dictionary with string keys, where the keys
 110     name command line "flags" or options.
 111
 112     Example:
 113         Execute a command named `exe` that takes a flagged option for file name
 114         (stored in a local Python variable `my_filename`) and an `origin` flag
 115         that uses the next three arguments to define a vector.
 116
 117             >>> my_filename = "somefilename"
 118             >>> result = cli(('exe', '--origin', 1.0, 2.0, 3.0, '-f', my_filename), shell=False)
 119             >>> assert hasattr(result, 'file')
 120             >>> assert hasattr(result, 'erroroutput')
 121             >>> assert hasattr(result, 'returncode')
 122
 123     Returns:
 124         A data structure with attributes for each of the results `file`, `erroroutput`, and `returncode`
 125
 126     Result object attributes:
 127         * `file`: the mapping of CLI flags to filename strings resulting from the `output` kwarg
 128         * `erroroutput`: A string of error output (if any) if the process failed.
 129         * `returncode`: return code of the subprocess.
 130
 131     """
 132     # Note: we could make provisions for stdio filehandles in a future version. E.g.
 133     # * STDOUT is available if a consuming operation is bound to `output.stdout`.
 134     # * STDERR is available if a consuming operation is bound to `output.stderr`.
 135     # * Otherwise, STDOUT and/or STDERR is(are) closed when command is called.
 136
 137     # In the operation implementation, we expect the `shell` parameter to be intercepted by the
 138     # wrapper and set to False.
 139     if shell:
 140         raise exceptions.UsageError("Operation does not support shell processing.")
 141
 142     if stdin == '':
 143         stdin = None
 144
 145     if isinstance(command, (str, bytes)):
 146         command = [command]
 147     command = list([arg for arg in command])
 148
 149     executable = shutil.which(command[0])
 150     if executable is None:
 151         raise exceptions.ValueError('"{}" is not found or not executable.'.format(command[0]))
 152     command[0] = executable
 153
 154     # TODO: (FR9) Can OS input/output filehandles be a responsibility of
 155     #  the code providing 'resources'?
 156
 157     erroroutput = ''
 158     logger.debug('executing subprocess')
 159     try:
 160         completed_process = subprocess.run(command,
 161                                            shell=shell,
 162                                            input=stdin,
 163                                            check=True,
 164                                            stdout=subprocess.PIPE,
 165                                            stderr=subprocess.STDOUT,
 166                                            universal_newlines=True
 167                                            )
 168         returncode = completed_process.returncode
 169         # TODO: Resource management code should manage a safe data object for `output`.
 170         for line in completed_process.stdout.split('\n'):
 171             logger.debug(line)
 172     except subprocess.CalledProcessError as e:
 173         logger.info("commandline operation had non-zero return status when calling {}".format(e.cmd))
 174         erroroutput = e.output
 175         returncode = e.returncode
 176     # Publish outputs.
 177     output.erroroutput = erroroutput
 178     output.returncode = returncode
 179
 180
 181 # TODO: (FR4) Make this a formal operation to properly handle gmxapi data dependencies.
 182 #  The consumer of this operation has an NDArray input. filemap may contain gmxapi data flow
 183 #  aspects that we want the framework to handle for us.
 184 def filemap_to_flag_list(filemap: dict = None):
 185     """Convert a map of command line flags and filenames to a list of command line arguments.
 186
 187     Used to map inputs and outputs of command line tools to and from gmxapi data handles.
 188     User provides mappings of flags and filenames so that gmxapi can construct an
 189     executable command line.
 190
 191     Primary use case is implicit. commandline_operation() instantiates this operation based on
 192     user input, and sends the output to cli()
 193
 194     Arguments:
 195         filemap: key-value map of command line flags and filename arguments
 196
 197     Returns:
 198         list of strings and/or gmxapi data references
 199     """
 200     result = []
 201     if filemap is not None:
 202         for key, value in filemap.items():
 203             # Note that the value may be a string, a list, an ndarray, or a future
 204             if not isinstance(value, (list, tuple, NDArray)):
 205                 if hasattr(value, 'result') and value.dtype == NDArray:
 206                     pass
 207                 elif hasattr(value, 'result') and value.dtype != NDArray:
 208                     # TODO: Fix this ugly hack when we have proper Future slicing and can make NDArray futures.
 209                     result_function = value.result
 210                     value.result = lambda function=result_function: [function()]
 211                 else:
 212                     value = [value]
 213             result = gmx.join_arrays(front=result, back=gmx.join_arrays(front=[key], back=value))
 214     return result
 215
 216
 217 # TODO: (FR4) Use generating function or decorator that can validate kwargs?
 218 # TODO: (FR4) Outputs need to be fully formed and typed in the object returned
 219 #  from the helper (decorated function).
 220 def commandline_operation(executable=None,
 221                           arguments=(),
 222                           input_files: dict = None,
 223                           output_files: dict = None,
 224                           stdin: str = None,
 225                           **kwargs):
 226     """Helper function to define a new operation that executes a subprocess in gmxapi data flow.
 227
 228     Define a new Operation for a particular executable and input/output parameter set.
 229     Generate a chain of operations to process the named key word arguments and handle
 230     input/output data dependencies.
 231
 232     Arguments:
 233         executable: name of an executable on the path
 234         arguments: list of positional arguments to insert at ``argv[1]``
 235         input_files: mapping of command-line flags to input file names
 236         output_files: mapping of command-line flags to output file names
 237         stdin (str): String input to send to STDIN (terminal input) of the executable (optional).
 238
 239     Multi-line text sent to *stdin* should be joined into a single string
 240     (e.g. ``'\n'.join(list_of_strings) + '\n'``).
 241     If multiple strings are provided to *stdin*, gmxapi will assume an ensemble,
 242     and will run one operation for each provided string.
 243
 244     Only string input (:py:func:str) to *stdin* is currently supported.
 245     If you have a use case that requires streaming input or binary input,
 246     please open an issue or contact the author(s).
 247
 248     Output:
 249         The output node of the resulting operation handle contains
 250         * ``file``: the mapping of CLI flags to filename strings resulting from the ``output_files`` kwarg
 251         * ``erroroutput``: A string of error output (if any) if the process failed.
 252         * ``returncode``: return code of the subprocess.
 253
 254     """
 255
 256     # Implementation details: When used in a script, this function returns an
 257     # instance of an operation. However, because of the dynamic specification of
 258     # inputs and outputs, each invocation may have the overhead of defining new
 259     # types to express the data flow topology, regardless of the executable.
 260     # If this overhead is problematic, consider exposing the intermediate step
 261     # at which the Operation is fully specified to facilitate reuse.
 262
 263     ##
 264     # 1. Define a new operation with outputs from `cli()` plus `file` from `output_files`
 265
 266     # output_files is essentially passed through, but we need assurance that results
 267     # will not be published until the rest of the operation has run (i.e. the cli() executable.)
 268
 269     # Warning: decorating a local function like this is counter to the notion of Operations
 270     # as portable (importable, serializable/deserializable). The big picture here needs
 271     # some more consideration.
 272     # TODO: (NOW) Distinguish portable Operations from relocatable Futures.
 273     # There is nothing antithetical about objects implementing gmxapi data interfaces
 274     # that are only resolvable by a certain Context as long as that Context can convey
 275     # the results to another Context upon request. Re-instantiating Operations is
 276     # only one way of relocating Futures. In this case, though, the dynamic creation of
 277     # merged_ops doesn't seem right, and commandline_operation should probably be
 278     # a proper Operation.
 279     #
 280     # TODO: (FR4+) Characterize the `file` dictionary key type:
 281     #  explicitly sequences rather than maybe-string/maybe-sequence-of-strings
 282     @gmx.function_wrapper(output={'erroroutput': str, 'returncode': int, 'file': dict})
 283     def merged_ops(erroroutput: str = None, returncode: int = None, file: dict = None,
 284                    output: OutputCollectionDescription = None):
 285         assert erroroutput is not None
 286         assert returncode is not None
 287         assert file is not None
 288         assert output is not None
 289         output.file = file
 290         output.returncode = returncode
 291         output.erroroutput = erroroutput
 292
 293     ##
 294     # 2. Prepare data flow.
 295
 296     if input_files is None:
 297         input_files = {}
 298     if output_files is None:
 299         output_files = {}
 300     if isinstance(arguments, (str, bytes)):
 301         arguments = [arguments]
 302     command = gmx.concatenate_lists([[executable],
 303                                      arguments,
 304                                      filemap_to_flag_list(input_files),
 305                                      filemap_to_flag_list(output_files)])
 306     shell = gmx.make_constant(False)
 307     cli_args = {'command': command,
 308                 'shell': shell}
 309     cli_args.update(**kwargs)
 310     if stdin is not None:
 311         cli_args['stdin'] = str(stdin)
 312
 313     ##
 314     # 3. Merge operations
 315     #
 316     # Note: Without a `label` argument, repeated calls to cli(**cli_args) should
 317     # produce references to the same unique resource. Creating this handle
 318     # separately should not be necessary, but we've got a way to go until we have the
 319     # fingerprinting and Context resource management we need for that.
 320     # TODO: ``label`` kwarg
 321     # TODO: input fingerprinting
 322     cli_result = cli(**cli_args)
 323     merged_result = merged_ops(erroroutput=cli_result.output.erroroutput,
 324                                returncode=cli_result.output.returncode,
 325                                file=output_files,
 326                                **kwargs)
 327
 328     # Return an object with an OutputCollection granting access to outputs of
 329     # cli() and of output_files (as "file")
 330     return merged_result