type inference doesn't work in parallel code sections

[robertwb]

The following example requires thread_id to be typed as int, even though the return type of parallel.threadid() is known.

def parallel_section():
    """
    >>> parallel_section()
    [1, 2, 3, 4, 5, 6, 7, 8, 9](0,)
    """
    #cdef int thread_id
    ids = []
    with nogil, parallel.parallel(num_threads=10):
        thread_id = parallel.threadid()   # id=0..9 for 10 threads
        with gil:
            ids.append( thread_id )
    return sorted(ids)

This is particularly annoying because it happens in a nogil section, so the code won't compile without explicitly typing the variable.

Migrated from http://trac.cython.org/ticket/774

[robertwb]

scoder changed description from

The following example requires thread_id to be typed as int, even though the return type of parallel.threadid() is known.

def parallel_section():
    """
    >>> parallel_section()
    [1, 2, 3, 4, 5, 6, 7, 8, 9](0,)
    """
    #cdef int thread_id
    ids = [   with nogil, parallel.parallel(num_threads=10):
        thread_id = parallel.threadid()   # id=0..9 for 10 threads
        with gil:
            ids.append( thread_id )
    return sorted(ids)

This is particularly annoying because it happens is a nogil section, so the code won't compile without explicitly typing the variable.

to

The following example requires thread_id to be typed as int, even though the return type of parallel.threadid() is known.

def parallel_section():
    """
    >>> parallel_section()
    [0, 1, 2, 3, 4, 5, 6, 7, 8, 9](]
)
    """
    #cdef int thread_id
    ids = []
    with nogil, parallel.parallel(num_threads=10):
        thread_id = parallel.threadid()   # id=0..9 for 10 threads
        with gil:
            ids.append( thread_id )
    return sorted(ids)

This is particularly annoying because it happens is a nogil section, so the code won't compile without explicitly typing the variable.
commented

[robertwb]

scoder changed description from

The following example requires thread_id to be typed as int, even though the return type of parallel.threadid() is known.

def parallel_section():
    """
    >>> parallel_section()
    [1, 2, 3, 4, 5, 6, 7, 8, 9](0,)
    """
    #cdef int thread_id
    ids = [   with nogil, parallel.parallel(num_threads=10):
        thread_id = parallel.threadid()   # id=0..9 for 10 threads
        with gil:
            ids.append( thread_id )
    return sorted(ids)

This is particularly annoying because it happens is a nogil section, so the code won't compile without explicitly typing the variable.

to

The following example requires thread_id to be typed as int, even though the return type of parallel.threadid() is known.

def parallel_section():
    """
    >>> parallel_section()
    [0, 1, 2, 3, 4, 5, 6, 7, 8, 9](]
)
    """
    #cdef int thread_id
    ids = []
    with nogil, parallel.parallel(num_threads=10):
        thread_id = parallel.threadid()   # id=0..9 for 10 threads
        with gil:
            ids.append( thread_id )
    return sorted(ids)

This is particularly annoying because it happens in a nogil section, so the code won't compile without explicitly typing the variable.
commented