xonsh/xonsh/execer.py
Anthony Scopatz c63410fba6 rm xonsherror
2016-05-14 01:51:47 -04:00

214 lines
9 KiB
Python

# -*- coding: utf-8 -*-
"""Implements the xonsh executer."""
import re
import types
import inspect
import builtins
from collections import Mapping
from xonsh import ast
from xonsh.parser import Parser
from xonsh.tools import subproc_toks, END_TOK_TYPES
from xonsh.built_ins import load_builtins, unload_builtins
RE_END_TOKS = re.compile('(;|and|\&\&|or|\|\||\))')
class Execer(object):
"""Executes xonsh code in a context."""
def __init__(self, filename='<xonsh-code>', debug_level=0, parser_args=None,
unload=True, config=None, login=True):
"""Parameters
----------
filename : str, optional
File we are to execute.
debug_level : int, optional
Debugging level to use in lexing and parsing.
parser_args : dict, optional
Arguments to pass down to the parser.
unload : bool, optional
Whether or not to unload xonsh builtins upon deletion.
config : str, optional
Path to configuration file.
"""
parser_args = parser_args or {}
self.parser = Parser(**parser_args)
self.filename = filename
self.debug_level = debug_level
self.unload = unload
self.ctxtransformer = ast.CtxAwareTransformer(self.parser)
load_builtins(execer=self, config=config, login=login)
def __del__(self):
if self.unload:
unload_builtins()
def parse(self, input, ctx, mode='exec'):
"""Parses xonsh code in a context-aware fashion. For context-free
parsing, please use the Parser class directly.
"""
if ctx is None:
ctx = set()
elif isinstance(ctx, Mapping):
ctx = set(ctx.keys())
# Parsing actually happens in a couple of phases. The first is a
# shortcut for a context-free parser. Normally, all subprocess
# lines should be wrapped in $(), to indicate that they are a
# subproc. But that would be super annoying. Unfortnately, Python
# mode - after indentation - is whitespace agnostic while, using
# the Python token, subproc mode is whitespace aware. That is to say,
# in Python mode "ls -l", "ls-l", and "ls - l" all parse to the
# same AST because whitespace doesn't matter to the minus binary op.
# However, these phases all have very different meaning in subproc
# mode. The 'right' way to deal with this is to make the entire
# grammar whitespace aware, and then ignore all of the whitespace
# tokens for all of the Python rules. The lazy way implemented here
# is to parse a line a second time with a $() wrapper if it fails
# the first time. This is a context-free phase.
tree, input = self._parse_ctx_free(input, mode=mode)
if tree is None:
return None
# Now we need to perform context-aware AST transformation. This is
# because the "ls -l" is valid Python. The only way that we know
# it is not actually Python is by checking to see if the first token
# (ls) is part of the execution context. If it isn't, then we will
# assume that this line is supposed to be a subprocess line, assuming
# it also is valid as a subprocess line.
tree = self.ctxtransformer.ctxvisit(tree, input, ctx, mode=mode)
return tree
def compile(self, input, mode='exec', glbs=None, locs=None, stacklevel=2,
filename=None):
"""Compiles xonsh code into a Python code object, which may then
be execed or evaled.
"""
if filename is None:
filename = self.filename
if glbs is None or locs is None:
frame = inspect.stack()[stacklevel][0]
glbs = frame.f_globals if glbs is None else glbs
locs = frame.f_locals if locs is None else locs
ctx = set(dir(builtins)) | set(glbs.keys()) | set(locs.keys())
tree = self.parse(input, ctx, mode=mode)
if tree is None:
return None # handles comment only input
code = compile(tree, filename, mode)
return code
def eval(self, input, glbs=None, locs=None, stacklevel=2):
"""Evaluates (and returns) xonsh code."""
if isinstance(input, types.CodeType):
code = input
else:
code = self.compile(input=input,
glbs=glbs,
locs=locs,
mode='eval',
stacklevel=stacklevel)
if code is None:
return None # handles comment only input
return eval(code, glbs, locs)
def exec(self, input, mode='exec', glbs=None, locs=None, stacklevel=2):
"""Execute xonsh code."""
if isinstance(input, types.CodeType):
code = input
else:
code = self.compile(input=input,
glbs=glbs,
locs=locs,
mode=mode,
stacklevel=stacklevel)
if code is None:
return None # handles comment only input
return exec(code, glbs, locs)
def _find_next_break(self, line, mincol):
if mincol >= 1:
line = line[mincol:]
if RE_END_TOKS.search(line) is None:
return None
maxcol = None
self.parser.lexer.input(line)
for tok in self.parser.lexer:
if tok.type in END_TOK_TYPES or \
(tok.type == 'ERRORTOKEN' and ')' in tok.value):
maxcol = tok.lexpos + mincol + 1
break
return maxcol
def _parse_ctx_free(self, input, mode='exec'):
last_error_line = last_error_col = -1
parsed = False
original_error = None
while not parsed:
try:
tree = self.parser.parse(input,
filename=self.filename,
mode=mode,
debug_level=self.debug_level)
parsed = True
except IndentationError as e:
if original_error is None:
raise e
else:
raise original_error
except SyntaxError as e:
if original_error is None:
original_error = e
if (e.loc is None) or (last_error_line == e.loc.lineno and
last_error_col in (e.loc.column + 1,
e.loc.column)):
raise original_error
last_error_col = e.loc.column
last_error_line = e.loc.lineno
idx = last_error_line - 1
lines = input.splitlines()
line = lines[idx]
if input.endswith('\n'):
lines.append('')
if len(line.strip()) == 0:
# whitespace only lines are not valid syntax in Python's
# interactive mode='single', who knew?! Just ignore them.
# this might cause actual sytax errors to have bad line
# numbers reported, but should only effect interactive mode
del lines[idx]
last_error_line = last_error_col = -1
input = '\n'.join(lines)
continue
if last_error_line > 1 and lines[idx-1].rstrip()[-1:] == ':':
# catch non-indented blocks and raise error.
prev_indent = len(lines[idx-1]) - len(lines[idx-1].lstrip())
curr_indent = len(lines[idx]) - len(lines[idx].lstrip())
if prev_indent == curr_indent:
raise original_error
maxcol = self._find_next_break(line, last_error_col)
sbpline = subproc_toks(line,
returnline=True,
maxcol=maxcol,
lexer=self.parser.lexer)
if sbpline is None:
# subprocess line had no valid tokens,
if len(line.partition('#')[0].strip()) == 0:
# likely because it only contained a comment.
del lines[idx]
last_error_line = last_error_col = -1
input = '\n'.join(lines)
continue
else:
# or for some other syntax error
raise original_error
elif sbpline.lstrip().startswith('![!['):
# if we have already wrapped this in subproc tokens
# and it still doesn't work, adding more won't help
# anything
raise original_error
else:
lines[idx] = sbpline
last_error_col += 3
input = '\n'.join(lines)
return tree, input