python create slice object from string

Question

I'd like to create a slice object from a string; right now the only way seems through a cumbersome hacky eval statement

class getslice:
    def __getitem__(self, idx): return idx[0]
eval("getslice()[%s, 1]" %(":-1"))

thanks in advance.

Edit: Sorry if the original prompt was not clear, the input in this case was ":-1". The point was to parse the string. Ignacio Vazquez-Abrams's response at least solved the problem (and seems to work with reverse indexing as well), but I think my solution above is still more clear if not conceptually clean (and will work correctly if Python ever changes slicing syntax).

Answer 1

I just needed to do this 12 years later so here's my answer using regex :)

import re
def parse_slice(string: str) -> slice:
    """
    Parse a string representation of a slice and return a slice object
    """
    # Matches one required colon, one optional colon, and up to three
    # positive or negative numbers between them
    match = re.match(r"^(-?[\d]*):(-?[\d]*)[:]?(-?[\d]*)$", string)
    if match:
        args = tuple(map(lambda s: int(s) if s else None, match.group(1, 2, 3)))
        return slice(*args)
    raise ValueError("Could not parse slice")

Answer 2

slice(*map(lambda x: int(x.strip()) if x.strip() else None, mystring.split(':')))

for single arg slices '-1' or '1' so when mystring.split(':')==1 you just call int(x)

On request, took it out of comment section.

Answer 3

My solution to parse numpy style advanced indexing from string: my gist. Although this is an old post, it's the only one I can find on this topic. Hope it helps.

Upon suggestion, I paste the code here, which could be a little bit long ... The code usage is (assuming a is an array-like object): a[parse_slice('1')] gives a[1]; a[parse_slice('2:,-1')] gives a[2:,-1]; etc.

import re

SLICE_TEMPLATES = [
    ('s', r'(?P<i>[+-]?\d+)'),
    ('sp', r'\((?P<i>[+-]?\d+)\)'),
    ('a', r'::?'),
    ('ri-', r'(?P<i>[+-]?\d+)::?'),
    ('ri-k', r'(?P<i>[+-]?\d+)::(?P<k>[+-]?\d+)'),
    ('r-j', r':(?P<j>[+-]?\d+):?'),
    ('r-jk', r':(?P<j>[+-]?\d+):(?P<k>[+-]?\d+)'),
    ('rij', r'(?P<i>[+-]?\d+):(?P<j>[+-]?\d+):?'),
    ('rijk', r'(?P<i>[+-]?\d+):(?P<j>[+-]?\d+):(?P<k>[+-]?\d+)'),
    ('r--k', r'::(?P<k>[+-]?\d+)'),
    ('l', r'\.\.\.'),
    ('eb', r'\[(?P<e>[+-]?\d+(,[+-]?\d+)*,?)\]'),
    ('ep', r'\((?P<e>[+-]?\d+(,[+-]?\d+)+,?)\)'),
    ('ep1', r'\((?P<e>[+-]?\d+,)\)'),
]
SLICE_TEMPLATES = [(k, re.compile(v)) for k, v in SLICE_TEMPLATES]


def tokenize_slice_groups(string):
    # tokenize groups
    groups = []
    sbuf = []
    expecting = {'(': ')', '[': ']'}
    pbbuf = []
    LEGAL_CHARS = '0123456789()[]+-:.'
    WHITESPACE_CHARS = ' \t'

    for c in string:
        if c in WHITESPACE_CHARS:
            pass
        elif c == ',':
            if len(pbbuf) not in (0, 2):
                sbuf.append(c)
            else:
                groups.append(''.join(sbuf))
                sbuf.clear()
                pbbuf.clear()
        elif c in LEGAL_CHARS:
            sbuf.append(c)
            if c in '([':
                if pbbuf:
                    raise ValueError('too many brackets in axis {}'.format(
                        len(groups)))
                pbbuf.append(c)
            elif c in ')]':
                if not pbbuf:
                    raise ValueError('brackets not match in axis {}'.format(
                        len(groups)))
                if c != expecting[pbbuf[0]]:
                    raise ValueError('brackets not match in axis {}'.format(
                        len(groups)))
                pbbuf.append(c)
        else:
            raise ValueError('illegal char `{}\''.format(c))
    groups.append(''.join(sbuf))
    return groups


def parse_slice_group(string):
    for name, tem in SLICE_TEMPLATES:
        matched = tem.fullmatch(string)
        if matched:
            if name[0] == 's':
                return int(matched.group('i'))
            if name[0] == 'a':
                return slice(None, None, None)
            if name[0] == 'r':
                i, j, k = None, None, None
                if 'i' in name:
                    i = int(matched.group('i'))
                if 'j' in name:
                    j = int(matched.group('j'))
                if 'k' in name:
                    k = int(matched.group('k'))
                return slice(i, j, k)
            if name[0] == 'l':
                return ...
            # if name[0] == 'e'
            return list(map(int, filter(None, matched.group('e').split(','))))
    raise ValueError('illegal group "{}"'.format(string))


def parse_slice(string):
    groups = tokenize_slice_groups(string)
    if groups == ['']:
        raise ValueError('index must not be empty')
    if groups and groups[-1] == '':
        del groups[-1]
    index = tuple(map(parse_slice_group, groups))
    if index.count(...) > 1:
        raise ValueError('ellipsis may occur at most once')
    return index

Answer 4

The one-liner from Ignacio Vazquez-Abrams is short but hardly readable and handles a single number inconsistently with slice. This tries to parse it in a cleaner way.

def parse_slice(value):
    """
    Parses a `slice()` from string, like `start:stop:step`.
    """
    if value:
        parts = value.split(':')
        if len(parts) == 1:
            # slice(stop)
            parts = [None, parts[0]]
        # else: slice(start, stop[, step])
    else:
        # slice()
        parts = []
    return slice(*[int(p) if p else None for p in parts])

# unit tests:
try:
    assert parse_slice('')
    assert False, 'It should raise TypeError'
except TypeError:
    pass
assert parse_slice('2') == slice(2)
assert parse_slice('2:3') == slice(2, 3)
assert parse_slice(':3') == slice(None, 3)
assert parse_slice(':') == slice(None, None)
assert parse_slice('2:') == slice(2, None)
assert parse_slice('2:3:4') == slice(2, 3, 4)
assert parse_slice(':3:4') == slice(None, 3, 4)
assert parse_slice('2::4') == slice(2, None, 4)
assert parse_slice('2:3:') == slice(2, 3, None)
assert parse_slice('::4') == slice(None, None, 4)
assert parse_slice('2::') == slice(2, None, None)
assert parse_slice('::') == slice(None, None, None)
assert parse_slice('-12:-13:-14') == slice(-12, -13, -14)
assert parse_slice('2:3:-4') == slice(2, 3, -4)
try:
    parse_slice('1:2:3:4')
    assert False, 'It should raise TypeError'
except TypeError:
    pass

Answer 5

How 'bout this (for simple non empty slice intervals) :

sliceStr = "3:8"
mySlice = slice( *map(int, sliceStr.split(':') ) )

Answer 6

Here's another method (just a consolidation of the others posted here):

def make_slice(expr):
    def to_piece(s):
        return s and int(s) or None
    pieces = map(to_piece, expr.split(':'))
    if len(pieces) == 1:
        return slice(pieces[0], pieces[0] + 1)
    else:
        return slice(*pieces)

Example usages:

In [1]: make_slice(':')
Out[1]: slice(None, None, None)

In [2]: make_slice(':-2')
Out[2]: slice(None, -2, None)

In [3]: x = [1, 2, 3]

In [4]: x[make_slice('::-1')]
Out[4]: [3, 2, 1]

Answer 7

Based on @pprzemak drafted the following function for elaborate parsing:

def parse_slice(v: Text):
    """
    Parses text like python "slice" expression (ie ``-10::2``).

    :param v:
        the slice expression or a lone integer
    :return:
        - None if input is None/empty
        - a ``slice()`` instance (even if input a lone numbrt)
    :raise ValueError:
        input non-empty but invalid syntax
    """
    orig_v = v
    v = v and v.strip()
    if not v:
        return

    try:
        if ':' not in v:
            ## A lone number given.
            v = int(v)
            return slice(v, v + 1)

        return slice(*map(lambda x: int(x.strip()) if x.strip() else None,
                          v.split(':')))
    except Exception:
        pass

    ## An alternative is to return `slice(None)` here.
    raise trt.TraitError("Syntax-error in '%s' slice!" % orig_v)

Answer 8

I end up here because I wanted my script to accept a python-like splice argument and render it into a list of integers, I did it with a function that seems like it answers the OP's question:

# create a slice object from a string
def get_slice_obj(slicearg):
    slice_ints = tuple([ int(n) for n in slicearg.split(':') ])
    return apply(slice, slice_ints)

def ints_from_slicearg(slicearg):
    slice_obj = get_slice_obj(slicearg)
    return(range(slice_obj.start or 0, slice_obj.stop or -1, slice_obj.step or 1))

for t in ['1', '1:3', '4:9:2']:
    print t, "=>", ints_from_slicearg(t)

Output:

1 => [0]
1:3 => [1, 2]
4:9:2 => [4, 6, 8]

Answer 9

A slice object is usually created using subscript notation, this notation uses slice() internally, as stated on the slice() documentation. What you want to do is:

your_string[start:end]

From the python tutorial:

Strings can be subscripted (indexed); like in C, the first character of a string has subscript (index) 0. There is no separate character type; a character is simply a string of size one. Like in Icon, substrings can be specified with the slice notation: two indices separated by a colon.

>>> word = 'Help' + 'A' 
>>> word[4]
'A'
>>> word[0:2]
'He'
>>> word[2:4]
'lp'

Slice indices have useful defaults; an omitted first index defaults to zero, an omitted second index defaults to the size of the string being sliced.

>>> word[:2]    # The first two characters
'He'
>>> word[2:]    # Everything except the first two characters
'lpA'

Answer 10

slice(*[{True: lambda n: None, False: int}[x == ''](x) for x in (mystring.split(':') + ['', '', ''])[:3]])

Answer 11

If you want a slice object, why don't you just instantiate one?

s = slice(start, stop, step)

What are you meaning by "creating it from a string"?