summaryrefslogtreecommitdiff
path: root/python/vyos/xml/definition.py
diff options
context:
space:
mode:
Diffstat (limited to 'python/vyos/xml/definition.py')
-rw-r--r--python/vyos/xml/definition.py294
1 files changed, 294 insertions, 0 deletions
diff --git a/python/vyos/xml/definition.py b/python/vyos/xml/definition.py
new file mode 100644
index 000000000..28a8e2a0c
--- /dev/null
+++ b/python/vyos/xml/definition.py
@@ -0,0 +1,294 @@
+# Copyright (C) 2020 VyOS maintainers and contributors
+#
+# This library is free software; you can redistribute it and/or modify it under the terms of
+# the GNU Lesser General Public License as published by the Free Software Foundation;
+# either version 2.1 of the License, or (at your option) any later version.
+#
+# This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY;
+# without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+# See the GNU Lesser General Public License for more details.
+#
+# You should have received a copy of the GNU Lesser General Public License along with this library;
+# if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+
+
+from vyos.xml import kw
+
+# As we index by key, the name is first and then the data:
+# {'dummy': {
+# '[node]': '[tagNode]',
+# 'address': { ... }
+# } }
+
+# so when we encounter a tagNode, we are really encountering
+# the tagNode data.
+
+
+class XML(dict):
+ def __init__(self):
+ self[kw.tree] = {}
+ self[kw.priorities] = {}
+ self[kw.owners] = {}
+ self[kw.default] = {}
+ self[kw.tags] = []
+
+ dict.__init__(self)
+
+ self.tree = self[kw.tree]
+ # the options which matched the last incomplete world we had
+ # or the last word in a list
+ self.options = []
+ # store all the part of the command we processed
+ self.inside = []
+ # should we check the data pass with the constraints
+ self.check = False
+ # are we still typing a word
+ self.filling = False
+ # do what have the tagNode value ?
+ self.filled = False
+ # last word seen
+ self.word = ''
+ # do we have all the data we want ?
+ self.final = False
+ # do we have too much data ?
+ self.extra = False
+ # what kind of node are we in plain vs data not
+ self.plain = True
+
+ def reset(self):
+ self.tree = self[kw.tree]
+ self.options = []
+ self.inside = []
+ self.check = False
+ self.filling = False
+ self.filled = False
+ self.word = ''
+ self.final = False
+ self.extra = False
+ self.plain = True
+
+ # from functools import lru_cache
+ # @lru_cache(maxsize=100)
+ # XXX: need to use cachetool instead - for later
+
+ def traverse(self, cmd):
+ self.reset()
+
+ # using split() intead of split(' ') eats the final ' '
+ words = cmd.split(' ')
+ passed = []
+ word = ''
+ data_node = False
+ space = False
+
+ while words:
+ word = words.pop(0)
+ space = word == ''
+ perfect = False
+ if word in self.tree:
+ passed = []
+ perfect = True
+ self.tree = self.tree[word]
+ data_node = self.tree[kw.node]
+ self.inside.append(word)
+ word = ''
+ continue
+ if word and data_node:
+ passed.append(word)
+
+ is_valueless = self.tree.get(kw.valueless, False)
+ is_leafNode = data_node == kw.leafNode
+ is_dataNode = data_node in (kw.leafNode, kw.tagNode)
+ named_options = [_ for _ in self.tree if not kw.found(_)]
+
+ if is_leafNode:
+ self.final = is_valueless or len(passed) > 0
+ self.extra = is_valueless and len(passed) > 0
+ self.check = len(passed) >= 1
+ else:
+ self.final = False
+ self.extra = False
+ self.check = len(passed) == 1 and not space
+
+ if self.final:
+ self.word = ' '.join(passed)
+ else:
+ self.word = word
+
+ if self.final:
+ self.filling = True
+ else:
+ self.filling = not perfect and bool(cmd and word != '')
+
+ self.filled = self.final or (is_dataNode and len(passed) > 0 and word == '')
+
+ if is_dataNode and len(passed) == 0:
+ self.options = []
+ elif word:
+ if data_node != kw.plainNode or len(passed) == 1:
+ self.options = [_ for _ in self.tree if _.startswith(word)]
+ else:
+ self.options = []
+ else:
+ self.options = named_options
+
+ self.plain = not is_dataNode
+
+ # self.debug()
+
+ return self.word
+
+ def speculate(self):
+ if len(self.options) == 1:
+ self.tree = self.tree[self.options[0]]
+ self.word = ''
+ if self.tree.get(kw.node,'') not in (kw.tagNode, kw.leafNode):
+ self.options = [_ for _ in self.tree if not kw.found(_)]
+
+ def checks(self, cmd):
+ # as we move thought the named node twice
+ # the first time we get the data with the node
+ # and the second with the pass parameters
+ xml = self[kw.tree]
+
+ words = cmd.split(' ')
+ send = True
+ last = []
+ while words:
+ word = words.pop(0)
+ if word in xml:
+ xml = xml[word]
+ send = True
+ last = []
+ continue
+ if xml[kw.node] in (kw.tagNode, kw.leafNode):
+ if kw.constraint in xml:
+ if send:
+ yield (word, xml[kw.constraint])
+ send = False
+ else:
+ last.append((word, None))
+ if len(last) >= 2:
+ yield last[0]
+
+ def summary(self):
+ yield ('enter', '[ summary ]', str(self.inside))
+
+ if kw.help not in self.tree:
+ yield ('skip', '[ summary ]', str(self.inside))
+ return
+
+ if self.filled:
+ return
+
+ yield('', '', '\nHelp:')
+
+ if kw.help in self.tree:
+ summary = self.tree[kw.help].get(kw.summary)
+ values = self.tree[kw.help].get(kw.valuehelp, [])
+ if summary:
+ yield(summary, '', '')
+ for value in values:
+ yield(value[kw.format], value[kw.description], '')
+
+ def constraint(self):
+ yield ('enter', '[ constraint ]', str(self.inside))
+
+ if kw.help in self.tree:
+ yield ('skip', '[ constraint ]', str(self.inside))
+ return
+ if kw.error not in self.tree:
+ yield ('skip', '[ constraint ]', str(self.inside))
+ return
+ if not self.word or self.filling:
+ yield ('skip', '[ constraint ]', str(self.inside))
+ return
+
+ yield('', '', '\nData Constraint:')
+
+ yield('', 'constraint', str(self.tree[kw.error]))
+
+ def listing(self):
+ yield ('enter', '[ listing ]', str(self.inside))
+
+ # only show the details when we passed the tagNode data
+ if not self.plain and not self.filled:
+ yield ('skip', '[ listing ]', str(self.inside))
+ return
+
+ yield('', '', '\nPossible completions:')
+
+ options = list(self.tree.keys())
+ options.sort()
+ for option in options:
+ if kw.found(option):
+ continue
+ if not option.startswith(self.word):
+ continue
+ inner = self.tree[option]
+ prefix = '+> ' if inner.get(kw.node, '') != kw.leafNode else ' '
+ if kw.help in inner:
+ h = inner[kw.help]
+ yield (prefix + option, h.get(kw.summary), '')
+
+ def debug(self):
+ print('------')
+ print("word '%s'" % self.word)
+ print("filling " + str(self.filling))
+ print("filled " + str(self.filled))
+ print("final " + str(self.final))
+ print("extra " + str(self.extra))
+ print("plain " + str(self.plain))
+ print("options " + str(self.options))
+
+ # from functools import lru_cache
+ # @lru_cache(maxsize=100)
+ # XXX: need to use cachetool instead - for later
+
+ def defaults(self, lpath):
+ d = self[kw.default]
+ for k in lpath:
+ d = d[k]
+ r = {}
+
+ def _flatten(prefix, d, r):
+ for k in d:
+ if isinstance(d[k],dict):
+ key = f'{k}_' if not prefix else f'{prefix}{k}_'
+ _flatten(key, d[k], r)
+ continue
+ key = prefix + k
+ r[key.replace('-','_')] = d[k]
+
+ _flatten('', d, r)
+ return r
+
+ # from functools import lru_cache
+ # @lru_cache(maxsize=100)
+ # XXX: need to use cachetool instead - for later
+
+ def _get(self, lpath, tag):
+ tree = self[kw.tree]
+ spath = lpath.copy()
+ while spath:
+ p = spath.pop(0)
+ if p not in tree:
+ return None
+ tree = tree[p]
+ if tree[kw.node] == kw.tagNode and spath:
+ spath.pop(0)
+ if tag not in tree:
+ print(f'not in tree {lpath} {tag}')
+ return tree.get(tag, None)
+
+ def is_multi(self, lpath):
+ return self._get(lpath, kw.multi) is True
+
+ def is_tag(self, lpath):
+ return self._get(lpath, kw.node) == kw.tagNode
+
+ def is_leaf(self, lpath):
+ return self._get(lpath, kw.node) == kw.leafNode
+
+ def exists(self, lpath):
+ return self._get(lpath, kw.node) is not None