Warning
This document is for an in-development version of Galaxy. You can alternatively view this page in the latest release if it exists or view the top of the latest release's documentation.
Source code for galaxy.tools.actions.history_imp_exp
import logging
import os
import tempfile
from galaxy.tools.actions import ToolAction
from galaxy.tools.imp_exp import JobExportHistoryArchiveWrapper
from galaxy.util.odict import odict
log = logging.getLogger(__name__)
[docs]class ImportHistoryToolAction(ToolAction):
    """Tool action used for importing a history to an archive. """
[docs]    def execute(self, tool, trans, incoming={}, set_output_hid=False, overwrite=True, history=None, **kwargs):
        #
        # Create job.
        #
        trans.check_user_activation()
        job = trans.app.model.Job()
        session = trans.get_galaxy_session()
        job.session_id = session and session.id
        if history:
            history_id = history.id
        elif trans.history:
            history_id = trans.history.id
        else:
            history_id = None
        job.history_id = history_id
        job.tool_id = tool.id
        job.user_id = trans.user.id
        start_job_state = job.state  # should be job.states.NEW
        job.state = job.states.WAITING  # we need to set job state to something other than NEW, or else when tracking jobs in db it will be picked up before we have added input / output parameters
        trans.sa_session.add(job)
        trans.sa_session.flush()  # ensure job.id are available
        #
        # Setup job and job wrapper.
        #
        # Add association for keeping track of job, history relationship.
        # Use abspath because mkdtemp() does not, contrary to the documentation,
        # always return an absolute path.
        archive_dir = os.path.abspath(tempfile.mkdtemp())
        jiha = trans.app.model.JobImportHistoryArchive(job=job, archive_dir=archive_dir)
        trans.sa_session.add(jiha)
        #
        # Add parameters to job_parameter table.
        #
        # Set additional parameters.
        incoming['__DEST_DIR__'] = jiha.archive_dir
        for name, value in tool.params_to_strings(incoming, trans.app).items():
            job.add_parameter(name, value)
        job.state = start_job_state  # job inputs have been configured, restore initial job state
        # Queue the job for execution
        trans.app.job_manager.enqueue(job, tool=tool)
        trans.log_event("Added import history job to the job queue, id: %s" % str(job.id), tool_id=job.tool_id)
        return job, odict()
[docs]class ExportHistoryToolAction(ToolAction):
    """Tool action used for exporting a history to an archive. """
[docs]    def execute(self, tool, trans, incoming={}, set_output_hid=False, overwrite=True, history=None, **kwargs):
        trans.check_user_activation()
        #
        # Get history to export.
        #
        history = None
        for name, value in incoming.items():
            if isinstance(value, trans.app.model.History):
                history_param_name = name
                history = value
                del incoming[history_param_name]
                break
        if not history:
            raise Exception('There is no history to export.')
        #
        # Create the job and output dataset objects
        #
        job = trans.app.model.Job()
        session = trans.get_galaxy_session()
        job.session_id = session and session.id
        if history:
            history_id = history.id
        else:
            history_id = trans.history.id
        job.history_id = history_id
        job.tool_id = tool.id
        if trans.user:
            # If this is an actual user, run the job as that individual.  Otherwise we're running as guest.
            job.user_id = trans.user.id
        start_job_state = job.state  # should be job.states.NEW
        job.state = job.states.WAITING  # we need to set job state to something other than NEW, or else when tracking jobs in db it will be picked up before we have added input / output parameters
        trans.sa_session.add(job)
        # Create dataset that will serve as archive.
        archive_dataset = trans.app.model.Dataset()
        trans.sa_session.add(archive_dataset)
        trans.sa_session.flush()  # ensure job.id and archive_dataset.id are available
        trans.app.object_store.create(archive_dataset)  # set the object store id, create dataset (if applicable)
        #
        # Setup job and job wrapper.
        #
        # Add association for keeping track of job, history, archive relationship.
        jeha = trans.app.model.JobExportHistoryArchive(job=job, history=history,
                                                       dataset=archive_dataset,
                                                       compressed=incoming['compress'])
        trans.sa_session.add(jeha)
        job_wrapper = JobExportHistoryArchiveWrapper(trans.app, job)
        cmd_line = job_wrapper.setup_job(jeha, include_hidden=incoming['include_hidden'],
                                         include_deleted=incoming['include_deleted'])
        #
        # Add parameters to job_parameter table.
        #
        # Set additional parameters.
        incoming['__HISTORY_TO_EXPORT__'] = history.id
        incoming['__EXPORT_HISTORY_COMMAND_INPUTS_OPTIONS__'] = cmd_line
        for name, value in tool.params_to_strings(incoming, trans.app).items():
            job.add_parameter(name, value)
        job.state = start_job_state  # job inputs have been configured, restore initial job state
        # Queue the job for execution
        trans.app.job_manager.enqueue(job, tool=tool)
        trans.log_event("Added export history job to the job queue, id: %s" % str(job.id), tool_id=job.tool_id)
        return job, odict()