Exemplo n.º 1
0
    def iter(self, iters, executor=None):
        """Iterate over dependencies and yield results

        :param iters: dict with iterators over dependencies
        :param executor: Executor to punt computation tasks to. If None,
            will compute inside the plugin's thread.
        """
        deps_by_kind = self.dependencies_by_kind()

        # Merge iterators of data that has the same kind
        kind_iters = dict()
        for kind, deps in deps_by_kind.items():
            kind_iters[kind] = strax.merge_iters(
                strax.sync_iters(strax.same_length,
                                 {d: iters[d]
                                  for d in deps}))

        if len(deps_by_kind) > 1:
            # Sync iterators of different kinds by time
            kind_iters = strax.sync_iters(
                partial(strax.same_stop, func=strax.endtime), kind_iters)

        iters = kind_iters
        pending = []
        yield from self._inner_iter(iters, pending, executor)
        self.cleanup(wait_for=pending)
Exemplo n.º 2
0
    def iter(self, iters, executor=None):
        """Iterate over dependencies and yield results

        :param iters: dict with iterators over dependencies
        :param executor: Executor to punt computation tasks to. If None,
            will compute inside the plugin's thread.
        """
        deps_by_kind = self.dependencies_by_kind()

        if len(deps_by_kind) > 1:
            # Sync the iterators that provide time info for each data kind
            # (first in deps_by_kind lists) by endtime
            iters.update(
                strax.sync_iters(
                    partial(strax.same_stop, func=strax.endtime),
                    {d[0]: iters[d[0]]
                     for d in deps_by_kind.values()}))

        # Convert to iterators over merged data for each kind
        new_iters = dict()
        for kind, deps in deps_by_kind.items():
            if len(deps) > 1:
                synced_iters = strax.sync_iters(strax.same_length,
                                                {d: iters[d]
                                                 for d in deps})
                new_iters[kind] = strax.merge_iters(synced_iters.values())
            else:
                new_iters[kind] = iters[deps[0]]
        iters = new_iters

        if self.rechunk_input:
            iters = self.rechunk_input(iters)

        pending = []
        for chunk_i in itertools.count():
            if not self.is_ready(chunk_i):
                if self.source_finished():
                    break
                print(f"{self.__class__.__name__} waiting for chunk {chunk_i}")
                time.sleep(2)

            try:
                compute_kwargs = {k: next(iters[k]) for k in deps_by_kind}
            except StopIteration:
                return
            except Exception:
                raise

            if self.parallel and executor is not None:
                new_f = executor.submit(self.do_compute,
                                        chunk_i=chunk_i,
                                        **compute_kwargs)
                pending = [f for f in pending + [new_f] if not f.done()]
                yield new_f
            else:
                yield self.do_compute(chunk_i=chunk_i, **compute_kwargs)

        self.cleanup(wait_for=pending)
Exemplo n.º 3
0
    def iter(self, iters, executor=None):
        """Iterate over dependencies and yield results
        :param iters: dict with iterators over dependencies
        :param executor: Executor to punt computation tasks to.
            If None, will compute inside the plugin's thread.
        """
        deps_by_kind = self.dependencies_by_kind()

        if len(deps_by_kind) > 1:
            # Sync the iterators that provide time info for each data kind
            # (first in deps_by_kind lists) by endtime
            iters.update(
                strax.sync_iters(
                    partial(strax.same_stop, func=strax.endtime),
                    {d[0]: iters[d[0]]
                     for d in deps_by_kind.values()}))

        # Convert to iterators over merged data for each kind
        new_iters = dict()
        for kind, deps in deps_by_kind.items():
            if len(deps) > 1:
                synced_iters = strax.sync_iters(strax.same_length,
                                                {d: iters[d]
                                                 for d in deps})
                new_iters[kind] = strax.merge_iters(synced_iters.values())
            else:
                new_iters[kind] = iters[deps[0]]
        iters = new_iters

        if self.rechunk_input:
            iters = self.rechunk_input(iters)

        pending = []
        for chunk_i in itertools.count():
            try:
                if not self.check_next_ready_or_done(chunk_i):
                    # TODO: avoid duplication
                    # but also remain picklable...
                    self.close(wait_for=tuple(pending))
                    return
                compute_kwargs = {k: next(iters[k]) for k in deps_by_kind}
            except StopIteration:
                self.close(wait_for=tuple(pending))
                return
            except Exception:
                self.close(wait_for=tuple(pending))
                raise

            if self.parallel and executor is not None:
                new_f = executor.submit(self.do_compute,
                                        chunk_i=chunk_i,
                                        **compute_kwargs)
                pending = [f for f in pending + [new_f] if not f.done()]
                yield new_f
            else:
                yield self.do_compute(chunk_i=chunk_i, **compute_kwargs)
Exemplo n.º 4
0
def test_sync_iters(source, source_skipper):
    synced = strax.sync_iters(strax.same_stop,
                              dict(s1=source, s2=source_skipper))
    assert len(synced) == 2
    assert 's1' in synced and 's2' in synced

    _do_sync_check(list(synced['s1']), list(synced['s2']))