forked from platformio/platformio-core
Move INO converter to a separate tool
This commit is contained in:
254
platformio/builder/tools/pioino.py
Normal file
254
platformio/builder/tools/pioino.py
Normal file
@ -0,0 +1,254 @@
|
||||
# Copyright (c) 2014-present PlatformIO <contact@platformio.org>
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
from __future__ import absolute_import
|
||||
|
||||
import atexit
|
||||
import glob
|
||||
import io
|
||||
import os
|
||||
import re
|
||||
import tempfile
|
||||
|
||||
import click
|
||||
|
||||
from platformio.compat import get_filesystem_encoding, get_locale_encoding
|
||||
|
||||
|
||||
class InoToCPPConverter(object):
|
||||
|
||||
PROTOTYPE_RE = re.compile(
|
||||
r"""^(
|
||||
(?:template\<.*\>\s*)? # template
|
||||
([a-z_\d\&]+\*?\s+){1,2} # return type
|
||||
([a-z_\d]+\s*) # name of prototype
|
||||
\([a-z_,\.\*\&\[\]\s\d]*\) # arguments
|
||||
)\s*(\{|;) # must end with `{` or `;`
|
||||
""",
|
||||
re.X | re.M | re.I,
|
||||
)
|
||||
DETECTMAIN_RE = re.compile(r"void\s+(setup|loop)\s*\(", re.M | re.I)
|
||||
PROTOPTRS_TPLRE = r"\([^&\(]*&(%s)[^\)]*\)"
|
||||
|
||||
def __init__(self, env):
|
||||
self.env = env
|
||||
self._main_ino = None
|
||||
self._safe_encoding = None
|
||||
|
||||
def read_safe_contents(self, path):
|
||||
error_reported = False
|
||||
for encoding in (
|
||||
"utf-8",
|
||||
None,
|
||||
get_filesystem_encoding(),
|
||||
get_locale_encoding(),
|
||||
"latin-1",
|
||||
):
|
||||
try:
|
||||
with io.open(path, encoding=encoding) as fp:
|
||||
contents = fp.read()
|
||||
self._safe_encoding = encoding
|
||||
return contents
|
||||
except UnicodeDecodeError:
|
||||
if not error_reported:
|
||||
error_reported = True
|
||||
click.secho(
|
||||
"Unicode decode error has occurred, please remove invalid "
|
||||
"(non-ASCII or non-UTF8) characters from %s file or convert it to UTF-8"
|
||||
% path,
|
||||
fg="yellow",
|
||||
err=True,
|
||||
)
|
||||
return ""
|
||||
|
||||
def write_safe_contents(self, path, contents):
|
||||
with io.open(
|
||||
path, "w", encoding=self._safe_encoding, errors="backslashreplace"
|
||||
) as fp:
|
||||
return fp.write(contents)
|
||||
|
||||
def is_main_node(self, contents):
|
||||
return self.DETECTMAIN_RE.search(contents)
|
||||
|
||||
def convert(self, nodes):
|
||||
contents = self.merge(nodes)
|
||||
if not contents:
|
||||
return None
|
||||
return self.process(contents)
|
||||
|
||||
def merge(self, nodes):
|
||||
assert nodes
|
||||
lines = []
|
||||
for node in nodes:
|
||||
contents = self.read_safe_contents(node.get_path())
|
||||
_lines = ['# 1 "%s"' % node.get_path().replace("\\", "/"), contents]
|
||||
if self.is_main_node(contents):
|
||||
lines = _lines + lines
|
||||
self._main_ino = node.get_path()
|
||||
else:
|
||||
lines.extend(_lines)
|
||||
|
||||
if not self._main_ino:
|
||||
self._main_ino = nodes[0].get_path()
|
||||
|
||||
return "\n".join(["#include <Arduino.h>"] + lines) if lines else None
|
||||
|
||||
def process(self, contents):
|
||||
out_file = self._main_ino + ".cpp"
|
||||
assert self._gcc_preprocess(contents, out_file)
|
||||
contents = self.read_safe_contents(out_file)
|
||||
contents = self._join_multiline_strings(contents)
|
||||
self.write_safe_contents(out_file, self.append_prototypes(contents))
|
||||
return out_file
|
||||
|
||||
def _gcc_preprocess(self, contents, out_file):
|
||||
tmp_path = tempfile.mkstemp()[1]
|
||||
self.write_safe_contents(tmp_path, contents)
|
||||
self.env.Execute(
|
||||
self.env.VerboseAction(
|
||||
'$CXX -o "{0}" -x c++ -fpreprocessed -dD -E "{1}"'.format(
|
||||
out_file, tmp_path
|
||||
),
|
||||
"Converting " + os.path.basename(out_file[:-4]),
|
||||
)
|
||||
)
|
||||
atexit.register(_delete_file, tmp_path)
|
||||
return os.path.isfile(out_file)
|
||||
|
||||
def _join_multiline_strings(self, contents):
|
||||
if "\\\n" not in contents:
|
||||
return contents
|
||||
newlines = []
|
||||
linenum = 0
|
||||
stropen = False
|
||||
for line in contents.split("\n"):
|
||||
_linenum = self._parse_preproc_line_num(line)
|
||||
if _linenum is not None:
|
||||
linenum = _linenum
|
||||
else:
|
||||
linenum += 1
|
||||
|
||||
if line.endswith("\\"):
|
||||
if line.startswith('"'):
|
||||
stropen = True
|
||||
newlines.append(line[:-1])
|
||||
continue
|
||||
if stropen:
|
||||
newlines[len(newlines) - 1] += line[:-1]
|
||||
continue
|
||||
elif stropen and line.endswith(('",', '";')):
|
||||
newlines[len(newlines) - 1] += line
|
||||
stropen = False
|
||||
newlines.append(
|
||||
'#line %d "%s"' % (linenum, self._main_ino.replace("\\", "/"))
|
||||
)
|
||||
continue
|
||||
|
||||
newlines.append(line)
|
||||
|
||||
return "\n".join(newlines)
|
||||
|
||||
@staticmethod
|
||||
def _parse_preproc_line_num(line):
|
||||
if not line.startswith("#"):
|
||||
return None
|
||||
tokens = line.split(" ", 3)
|
||||
if len(tokens) > 2 and tokens[1].isdigit():
|
||||
return int(tokens[1])
|
||||
return None
|
||||
|
||||
def _parse_prototypes(self, contents):
|
||||
prototypes = []
|
||||
reserved_keywords = set(["if", "else", "while"])
|
||||
for match in self.PROTOTYPE_RE.finditer(contents):
|
||||
if (
|
||||
set([match.group(2).strip(), match.group(3).strip()])
|
||||
& reserved_keywords
|
||||
):
|
||||
continue
|
||||
prototypes.append(match)
|
||||
return prototypes
|
||||
|
||||
def _get_total_lines(self, contents):
|
||||
total = 0
|
||||
if contents.endswith("\n"):
|
||||
contents = contents[:-1]
|
||||
for line in contents.split("\n")[::-1]:
|
||||
linenum = self._parse_preproc_line_num(line)
|
||||
if linenum is not None:
|
||||
return total + linenum
|
||||
total += 1
|
||||
return total
|
||||
|
||||
def append_prototypes(self, contents):
|
||||
prototypes = self._parse_prototypes(contents) or []
|
||||
|
||||
# skip already declared prototypes
|
||||
declared = set(m.group(1).strip() for m in prototypes if m.group(4) == ";")
|
||||
prototypes = [m for m in prototypes if m.group(1).strip() not in declared]
|
||||
|
||||
if not prototypes:
|
||||
return contents
|
||||
|
||||
prototype_names = set(m.group(3).strip() for m in prototypes)
|
||||
split_pos = prototypes[0].start()
|
||||
match_ptrs = re.search(
|
||||
self.PROTOPTRS_TPLRE % ("|".join(prototype_names)),
|
||||
contents[:split_pos],
|
||||
re.M,
|
||||
)
|
||||
if match_ptrs:
|
||||
split_pos = contents.rfind("\n", 0, match_ptrs.start()) + 1
|
||||
|
||||
result = []
|
||||
result.append(contents[:split_pos].strip())
|
||||
result.append("%s;" % ";\n".join([m.group(1) for m in prototypes]))
|
||||
result.append(
|
||||
'#line %d "%s"'
|
||||
% (
|
||||
self._get_total_lines(contents[:split_pos]),
|
||||
self._main_ino.replace("\\", "/"),
|
||||
)
|
||||
)
|
||||
result.append(contents[split_pos:].strip())
|
||||
return "\n".join(result)
|
||||
|
||||
|
||||
def ConvertInoToCpp(env):
|
||||
src_dir = glob.escape(env.subst("$PROJECT_SRC_DIR"))
|
||||
ino_nodes = env.Glob(os.path.join(src_dir, "*.ino")) + env.Glob(
|
||||
os.path.join(src_dir, "*.pde")
|
||||
)
|
||||
if not ino_nodes:
|
||||
return
|
||||
c = InoToCPPConverter(env)
|
||||
out_file = c.convert(ino_nodes)
|
||||
|
||||
atexit.register(_delete_file, out_file)
|
||||
|
||||
|
||||
def _delete_file(path):
|
||||
try:
|
||||
if os.path.isfile(path):
|
||||
os.remove(path)
|
||||
except: # pylint: disable=bare-except
|
||||
pass
|
||||
|
||||
|
||||
def generate(env):
|
||||
env.AddMethod(ConvertInoToCpp)
|
||||
|
||||
|
||||
def exists(_):
|
||||
return True
|
@ -14,244 +14,15 @@
|
||||
|
||||
from __future__ import absolute_import
|
||||
|
||||
import atexit
|
||||
import glob
|
||||
import io
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
import tempfile
|
||||
|
||||
import click
|
||||
|
||||
from platformio import fs, util
|
||||
from platformio.compat import get_filesystem_encoding, get_locale_encoding
|
||||
from platformio.package.manager.core import get_core_package_dir
|
||||
from platformio.proc import exec_command
|
||||
|
||||
|
||||
class InoToCPPConverter(object):
|
||||
|
||||
PROTOTYPE_RE = re.compile(
|
||||
r"""^(
|
||||
(?:template\<.*\>\s*)? # template
|
||||
([a-z_\d\&]+\*?\s+){1,2} # return type
|
||||
([a-z_\d]+\s*) # name of prototype
|
||||
\([a-z_,\.\*\&\[\]\s\d]*\) # arguments
|
||||
)\s*(\{|;) # must end with `{` or `;`
|
||||
""",
|
||||
re.X | re.M | re.I,
|
||||
)
|
||||
DETECTMAIN_RE = re.compile(r"void\s+(setup|loop)\s*\(", re.M | re.I)
|
||||
PROTOPTRS_TPLRE = r"\([^&\(]*&(%s)[^\)]*\)"
|
||||
|
||||
def __init__(self, env):
|
||||
self.env = env
|
||||
self._main_ino = None
|
||||
self._safe_encoding = None
|
||||
|
||||
def read_safe_contents(self, path):
|
||||
error_reported = False
|
||||
for encoding in (
|
||||
"utf-8",
|
||||
None,
|
||||
get_filesystem_encoding(),
|
||||
get_locale_encoding(),
|
||||
"latin-1",
|
||||
):
|
||||
try:
|
||||
with io.open(path, encoding=encoding) as fp:
|
||||
contents = fp.read()
|
||||
self._safe_encoding = encoding
|
||||
return contents
|
||||
except UnicodeDecodeError:
|
||||
if not error_reported:
|
||||
error_reported = True
|
||||
click.secho(
|
||||
"Unicode decode error has occurred, please remove invalid "
|
||||
"(non-ASCII or non-UTF8) characters from %s file or convert it to UTF-8"
|
||||
% path,
|
||||
fg="yellow",
|
||||
err=True,
|
||||
)
|
||||
return ""
|
||||
|
||||
def write_safe_contents(self, path, contents):
|
||||
with io.open(
|
||||
path, "w", encoding=self._safe_encoding, errors="backslashreplace"
|
||||
) as fp:
|
||||
return fp.write(contents)
|
||||
|
||||
def is_main_node(self, contents):
|
||||
return self.DETECTMAIN_RE.search(contents)
|
||||
|
||||
def convert(self, nodes):
|
||||
contents = self.merge(nodes)
|
||||
if not contents:
|
||||
return None
|
||||
return self.process(contents)
|
||||
|
||||
def merge(self, nodes):
|
||||
assert nodes
|
||||
lines = []
|
||||
for node in nodes:
|
||||
contents = self.read_safe_contents(node.get_path())
|
||||
_lines = ['# 1 "%s"' % node.get_path().replace("\\", "/"), contents]
|
||||
if self.is_main_node(contents):
|
||||
lines = _lines + lines
|
||||
self._main_ino = node.get_path()
|
||||
else:
|
||||
lines.extend(_lines)
|
||||
|
||||
if not self._main_ino:
|
||||
self._main_ino = nodes[0].get_path()
|
||||
|
||||
return "\n".join(["#include <Arduino.h>"] + lines) if lines else None
|
||||
|
||||
def process(self, contents):
|
||||
out_file = self._main_ino + ".cpp"
|
||||
assert self._gcc_preprocess(contents, out_file)
|
||||
contents = self.read_safe_contents(out_file)
|
||||
contents = self._join_multiline_strings(contents)
|
||||
self.write_safe_contents(out_file, self.append_prototypes(contents))
|
||||
return out_file
|
||||
|
||||
def _gcc_preprocess(self, contents, out_file):
|
||||
tmp_path = tempfile.mkstemp()[1]
|
||||
self.write_safe_contents(tmp_path, contents)
|
||||
self.env.Execute(
|
||||
self.env.VerboseAction(
|
||||
'$CXX -o "{0}" -x c++ -fpreprocessed -dD -E "{1}"'.format(
|
||||
out_file, tmp_path
|
||||
),
|
||||
"Converting " + os.path.basename(out_file[:-4]),
|
||||
)
|
||||
)
|
||||
atexit.register(_delete_file, tmp_path)
|
||||
return os.path.isfile(out_file)
|
||||
|
||||
def _join_multiline_strings(self, contents):
|
||||
if "\\\n" not in contents:
|
||||
return contents
|
||||
newlines = []
|
||||
linenum = 0
|
||||
stropen = False
|
||||
for line in contents.split("\n"):
|
||||
_linenum = self._parse_preproc_line_num(line)
|
||||
if _linenum is not None:
|
||||
linenum = _linenum
|
||||
else:
|
||||
linenum += 1
|
||||
|
||||
if line.endswith("\\"):
|
||||
if line.startswith('"'):
|
||||
stropen = True
|
||||
newlines.append(line[:-1])
|
||||
continue
|
||||
if stropen:
|
||||
newlines[len(newlines) - 1] += line[:-1]
|
||||
continue
|
||||
elif stropen and line.endswith(('",', '";')):
|
||||
newlines[len(newlines) - 1] += line
|
||||
stropen = False
|
||||
newlines.append(
|
||||
'#line %d "%s"' % (linenum, self._main_ino.replace("\\", "/"))
|
||||
)
|
||||
continue
|
||||
|
||||
newlines.append(line)
|
||||
|
||||
return "\n".join(newlines)
|
||||
|
||||
@staticmethod
|
||||
def _parse_preproc_line_num(line):
|
||||
if not line.startswith("#"):
|
||||
return None
|
||||
tokens = line.split(" ", 3)
|
||||
if len(tokens) > 2 and tokens[1].isdigit():
|
||||
return int(tokens[1])
|
||||
return None
|
||||
|
||||
def _parse_prototypes(self, contents):
|
||||
prototypes = []
|
||||
reserved_keywords = set(["if", "else", "while"])
|
||||
for match in self.PROTOTYPE_RE.finditer(contents):
|
||||
if (
|
||||
set([match.group(2).strip(), match.group(3).strip()])
|
||||
& reserved_keywords
|
||||
):
|
||||
continue
|
||||
prototypes.append(match)
|
||||
return prototypes
|
||||
|
||||
def _get_total_lines(self, contents):
|
||||
total = 0
|
||||
if contents.endswith("\n"):
|
||||
contents = contents[:-1]
|
||||
for line in contents.split("\n")[::-1]:
|
||||
linenum = self._parse_preproc_line_num(line)
|
||||
if linenum is not None:
|
||||
return total + linenum
|
||||
total += 1
|
||||
return total
|
||||
|
||||
def append_prototypes(self, contents):
|
||||
prototypes = self._parse_prototypes(contents) or []
|
||||
|
||||
# skip already declared prototypes
|
||||
declared = set(m.group(1).strip() for m in prototypes if m.group(4) == ";")
|
||||
prototypes = [m for m in prototypes if m.group(1).strip() not in declared]
|
||||
|
||||
if not prototypes:
|
||||
return contents
|
||||
|
||||
prototype_names = set(m.group(3).strip() for m in prototypes)
|
||||
split_pos = prototypes[0].start()
|
||||
match_ptrs = re.search(
|
||||
self.PROTOPTRS_TPLRE % ("|".join(prototype_names)),
|
||||
contents[:split_pos],
|
||||
re.M,
|
||||
)
|
||||
if match_ptrs:
|
||||
split_pos = contents.rfind("\n", 0, match_ptrs.start()) + 1
|
||||
|
||||
result = []
|
||||
result.append(contents[:split_pos].strip())
|
||||
result.append("%s;" % ";\n".join([m.group(1) for m in prototypes]))
|
||||
result.append(
|
||||
'#line %d "%s"'
|
||||
% (
|
||||
self._get_total_lines(contents[:split_pos]),
|
||||
self._main_ino.replace("\\", "/"),
|
||||
)
|
||||
)
|
||||
result.append(contents[split_pos:].strip())
|
||||
return "\n".join(result)
|
||||
|
||||
|
||||
def ConvertInoToCpp(env):
|
||||
src_dir = glob.escape(env.subst("$PROJECT_SRC_DIR"))
|
||||
ino_nodes = env.Glob(os.path.join(src_dir, "*.ino")) + env.Glob(
|
||||
os.path.join(src_dir, "*.pde")
|
||||
)
|
||||
if not ino_nodes:
|
||||
return
|
||||
c = InoToCPPConverter(env)
|
||||
out_file = c.convert(ino_nodes)
|
||||
|
||||
atexit.register(_delete_file, out_file)
|
||||
|
||||
|
||||
def _delete_file(path):
|
||||
try:
|
||||
if os.path.isfile(path):
|
||||
os.remove(path)
|
||||
except: # pylint: disable=bare-except
|
||||
pass
|
||||
|
||||
|
||||
@util.memoized()
|
||||
def _get_compiler_type(env):
|
||||
def GetCompilerType(env):
|
||||
if env.subst("$CC").endswith("-gcc"):
|
||||
return "gcc"
|
||||
try:
|
||||
@ -270,10 +41,6 @@ def _get_compiler_type(env):
|
||||
return None
|
||||
|
||||
|
||||
def GetCompilerType(env):
|
||||
return _get_compiler_type(env)
|
||||
|
||||
|
||||
def GetActualLDScript(env):
|
||||
def _lookup_in_ldpath(script):
|
||||
for d in env.get("LIBPATH", []):
|
||||
@ -319,7 +86,7 @@ def GetActualLDScript(env):
|
||||
env.Exit(1)
|
||||
|
||||
|
||||
def ConfigureDebugFlags(env):
|
||||
def ConfigureDebugTarget(env):
|
||||
def _cleanup_debug_flags(scope):
|
||||
if scope not in env:
|
||||
return
|
||||
@ -384,7 +151,6 @@ def exists(_):
|
||||
|
||||
|
||||
def generate(env):
|
||||
env.AddMethod(ConvertInoToCpp)
|
||||
env.AddMethod(GetCompilerType)
|
||||
env.AddMethod(GetActualLDScript)
|
||||
env.AddMethod(ConfigureDebugFlags)
|
||||
|
Reference in New Issue
Block a user