magic slurm node list parser

nodelist.py

#!/usr/bin/env python
import re

node_lists = [
  "clip-g1-[0-1],clip-g2-[2-3]",
  "clip-g1-0,clip-g2-0",
  "clip-g1-0,clip-g2-1",
  "clip-g1-1",
  "clip-a-[1,3,5]",
  "clip-b-[1-3,5]",
  "clip-c-[1-3,5,9-12]",
  "clip-d-[5,9-12]",
  "clip-e-[5,9],clip-e-[15-19]",
  "clip-f-[5,9],clip-f-[15,17]",
  "clip-f-5,clip-f-[15,17]",
  "clip-f-[5,9],clip-f-175"
]


# we assume index parts are sane
def expand_idx(idx):
  if idx.startswith('['):
    idx = idx[1:-1]
  idx_parts = idx.split(',')
  indexes = []
  for part in idx_parts:
    if "-" in part:
      start, stop = part.split('-', 2)
      indexes.extend(range(int(start), int(stop)+1))
    else:
      indexes.append(int(part))
  indexes.sort()
  return indexes

def expand_nodestr(nodestri, idx_list):
  return [nodestri + str(idx) for idx in idx_list]

index_pat = re.compile(r'((clip-(\w+)-)(\d+|\[[\d\,\-]+\]),?)')
for node_str in node_lists:
  print node_str
  for match in index_pat.finditer(node_str):
      print "  ", match.groups()
      #full = match.group(1)
      prefix = match.group(2)
      node_type = match.group(3)
      idx = match.group(4)
      idx_list = expand_idx(idx)
      print "prefix> ", prefix
      print "type>   ", node_type
      print "idx>    ", idx
      print "count>  ", len(idx_list)
      print "list>   ", expand_nodestr(prefix, idx_list)
~                                                              

regex support by @rembart. could not have done this without him

10 more lines, but skips regexp, and adds error checking:

node_lists = [
  "clip-g1-[0-1],clip-g2-[2-3]",
  "clip-g1-0,clip-g2-0",
  "clip-g1-0,clip-g2-1",
  "clip-g1-1",
  "clip-a-[1,3,5]",
  "clip-b-[1-3,5]",
  "clip-c-[1-3,5,9-12]",
  "clip-d-[5,9-12]",
  "clip-e-[5,9],clip-e-[15-19]",
  "clip-f-[5,9],clip-f-[15,17]",
  "clip-f-5,clip-f-[15,17]",
  "clip-f-[5,9],clip-f-175"
]

def parse_int(s):
    for i,c in enumerate(s):
        if c not in "0123456789":
            return int(s[:i]), s[i:]
    return int(s), ""

def parse_brackets(s):
    # parse a "bracket" expression (including closing ']')
    lst = []
    while len(s) > 0:
        if s[0] == ',':
            s = s[1:]
            continue
        if s[0] == ']':
            return lst, s[1:]
        a, s = parse_int(s)
        assert len(s) > 0, f"Missing closing ']'"
        if s[0] in ',]':
            lst.append(a)
        elif s[0] == '-':
            b, s = parse_int(s[1:])
            lst.extend(range(a,b+1))
    assert len(s) > 0, f"Missing closing ']'"

def parse_node(s):
    # parse a "node" expression
    for i,c in enumerate(s):
        if c == ',': # name,...
            return [ s[:i] ], s[i+1:]
        if c == '[': # name[v],...
            b, rest = parse_brackets(s[i+1:])
            if len(rest) > 0:
                assert rest[0] == ',', f"Expected comma after brackets in {s[i:]}"
                rest = rest[1:]
            return [s[:i]+str(z) for z in b], rest

    return [ s ], ""

def parse_list(s):
    lst = []
    while len(s) > 0:
        v, s = parse_node(s)
        lst.extend(v)
    return lst

for s in node_lists:
    print(s)
    print(parse_list(s))

For naming patterns with padding with zeros, such as cl001.

node_lists = [
  "clip-g1-[0-1],clip-g2-[2-3]",
  "clip-g1-0,clip-g2-0",
  "clip-g1-0,clip-g2-1",
  "clip-g1-1",
  "clip-a-[1,3,5]",
  "clip-b-[1-3,5]",
  "clip-c-[1-3,5,9-12]",
  "clip-d-[5,9-12]",
  "clip-e-[5,9],clip-e-[15-19]",
  "clip-f-[5,9],clip-f-[15,17]",
  "clip-f-5,clip-f-[15,17]",
  "clip-f-[5,9],clip-f-175",
  "cg[1-2]",
  "cg[001-002]",
  "cl[028,044,054]",
  "cg[001-002],cl[001-003,005-006,009-046]",
]

def parse_int(s):
    for i, c in enumerate(s):
        if c not in "0123456789":
            return s[:i], s[i:]
    return s, ""

def parse_brackets(s):
    # parse a "bracket" expression (including closing ']')
    lst = []
    while len(s) > 0:
        if s[0] == ',':
            s = s[1:]
            continue
        if s[0] == ']':
            return lst, s[1:]
        a, s = parse_int(s)
        assert len(s) > 0, "Missing closing ']'"
        if s[0] in ',]':
            lst.append(a)
        elif s[0] == '-':
            b, s = parse_int(s[1:])
            assert int(a) <= int(b), 'Invalid range'
            # A leading zero on a lower boundary suggests that the
            # numerical part of the node name is padded with zeros,
            # e.g. nia0001.
            #
            # Just a single 0 on the lower boundary suggests a numerical
            # range without padding, e.g. nia[0-4].
            if a != '0' and a.startswith('0'):
                assert len(a) == len(b), \
                    (
                        'Boundaries of a ranged string with padding '
                        'must have the same length.'
                    )
                lst.extend(
                    [str(x).zfill(len(a)) for x in range(int(a), int(b)+1)]
                )
            elif a != '0' and b.startswith('0'):
                raise ValueError('Could not determine the padding style.')
            # If no padding is detected, simply use the range.
            else:
                lst.extend(
                    [str(x) for x in range(int(a), int(b)+1)]
                )
    assert len(s) > 0, "Missing closing ']'"

def parse_node(s):
    # parse a "node" expression
    for i, c in enumerate(s):
        if c == ',':  # name,...
            return [s[:i]], s[i+1:]
        if c == '[':  # name[v],...
            b, rest = parse_brackets(s[i+1:])
            if len(rest) > 0:
                assert rest[0] == ',', \
                    f"Expected comma after brackets in {s[i:]}"
                rest = rest[1:]
            return [s[:i] + z for z in b], rest

    return [s], ""

def parse_list(s):
    lst = []
    while len(s) > 0:
        v, s = parse_node(s)
        lst.extend(v)
    return lst

for s in node_lists:
    print(s)
    print(parse_list(s))

… using EBNF grammar to parse - https://github.com/commonism/slurm_node_list_parser