summaryrefslogtreecommitdiff
path: root/gerber/gerber.py
diff options
context:
space:
mode:
Diffstat (limited to 'gerber/gerber.py')
-rw-r--r--gerber/gerber.py305
1 files changed, 305 insertions, 0 deletions
diff --git a/gerber/gerber.py b/gerber/gerber.py
new file mode 100644
index 0000000..31d9b82
--- /dev/null
+++ b/gerber/gerber.py
@@ -0,0 +1,305 @@
+#! /usr/bin/env python
+# -*- coding: utf-8 -*-
+
+# copyright 2014 Hamilton Kibbe <ham@hamiltonkib.be>
+# Modified from parser.py by Paulo Henrique Silva <ph.silva@gmail.com>
+#
+# 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.
+
+import re
+import json
+from .statements import *
+
+
+
+
+class GerberFile(object):
+ """ A class representing a single gerber file
+
+ The GerberFile class represents a single gerber file.
+
+ Parameters
+ ----------
+ filename : string
+ Parameter.
+
+ zero_suppression : string
+ Zero-suppression mode. May be either 'leading' or 'trailing'
+
+ notation : string
+ Notation mode. May be either 'absolute' or 'incremental'
+
+ format : tuple (int, int)
+ Gerber precision format expressed as a tuple containing:
+ (number of integer-part digits, number of decimal-part digits)
+
+ Attributes
+ ----------
+ comments: list of strings
+ List of comments contained in the gerber file.
+
+ units : string
+ either 'inch' or 'metric'.
+
+ size : tuple, (<float>, <float>)
+ Size in [self.units] of the layer described by the gerber file.
+
+ bounds: tuple, ((<float>, <float>), (<float>, <float>))
+ boundaries of the layer described by the gerber file.
+ `bounds` is stored as ((min x, max x), (min y, max y))
+
+ """
+
+ @classmethod
+ def read(cls, filename):
+ """ Read data from filename and return a GerberFile
+ """
+ return GerberParser().parse(filename)
+
+ def __init__(self, statements, settings, filename=None):
+ self.filename = filename
+ self.statements = statements
+ self.settings = settings
+
+ @property
+ def comments(self):
+ return [comment.comment for comment in self.statements
+ if isinstance(comment, CommentStmt)]
+
+ @property
+ def units(self):
+ return self.settings['units']
+
+ @property
+ def size(self):
+ xbounds, ybounds = self.bounds
+ return (xbounds[1] - xbounds[0], ybounds[1] - ybounds[0])
+
+ @property
+ def bounds(self):
+ xbounds = [0.0, 0.0]
+ ybounds = [0.0, 0.0]
+ for stmt in [stmt for stmt in self.statements if isinstance(stmt, CoordStmt)]:
+ if stmt.x is not None and stmt.x < xbounds[0]:
+ xbounds[0] = stmt.x
+ if stmt.x is not None and stmt.x > xbounds[1]:
+ xbounds[1] = stmt.x
+ if stmt.i is not None and stmt.i < xbounds[0]:
+ xbounds[0] = stmt.i
+ if stmt.i is not None and stmt.i > xbounds[1]:
+ xbounds[1] = stmt.i
+ if stmt.y is not None and stmt.y < ybounds[0]:
+ ybounds[0] = stmt.y
+ if stmt.y is not None and stmt.y > ybounds[1]:
+ ybounds[1] = stmt.y
+ if stmt.j is not None and stmt.j < ybounds[0]:
+ ybounds[0] = stmt.j
+ if stmt.j is not None and stmt.j > ybounds[1]:
+ ybounds[1] = stmt.j
+
+ return (xbounds, ybounds)
+
+ def write(self, filename):
+ """ Write data out to a gerber file
+ """
+ with open(filename, 'w') as f:
+ for statement in self.statements:
+ f.write(statement.to_gerber())
+
+ def render(self, filename, ctx):
+ """ Generate image of layer.
+ """
+ ctx.set_bounds(self.bounds)
+ for statement in self.statements:
+ ctx.evaluate(statement)
+ ctx.dump(filename)
+
+
+
+class GerberParser(object):
+ """ GerberParser
+ """
+ NUMBER = r"[\+-]?\d+"
+ DECIMAL = r"[\+-]?\d+([.]?\d+)?"
+ STRING = r"[a-zA-Z0-9_+\-/!?<>”’(){}.\|&@# :]+"
+ NAME = "[a-zA-Z_$][a-zA-Z_$0-9]+"
+ FUNCTION = r"G\d{2}"
+
+ COORD_OP = r"D[0]?[123]"
+
+ FS = r"(?P<param>FS)(?P<zero>(L|T))?(?P<notation>(A|I))X(?P<x>[0-7][0-7])Y(?P<y>[0-7][0-7])"
+ MO = r"(?P<param>MO)(?P<mo>(MM|IN))"
+ IP = r"(?P<param>IP)(?P<ip>(POS|NEG))"
+ LP = r"(?P<param>LP)(?P<lp>(D|C))"
+ AD_CIRCLE = r"(?P<param>AD)D(?P<d>\d+)(?P<shape>C)[,](?P<modifiers>[^,]*)"
+ AD_RECT = r"(?P<param>AD)D(?P<d>\d+)(?P<shape>R)[,](?P<modifiers>[^,]*)"
+ AD_OBROUND = r"(?P<param>AD)D(?P<d>\d+)(?P<shape>O)[,](?P<modifiers>[^,]*)"
+ AD_POLY = r"(?P<param>AD)D(?P<d>\d+)(?P<shape>P)[,](?P<modifiers>[^,]*)"
+ AD_MACRO = r"(?P<param>AD)D(?P<d>\d+)+(?P<shape>{name})[,](?P<modifiers>[^,]*)".format(name=NAME)
+ AM = r"(?P<param>AM)(?P<name>{name})\*(?P<macro>.*)".format(name=NAME)
+
+ # begin deprecated
+ OF = r"(?P<param>OF)(A(?P<a>{decimal}))?(B(?P<b>{decimal}))?".format(decimal=DECIMAL)
+ IN = r"(?P<param>IN)(?P<name>.*)"
+ LN = r"(?P<param>LN)(?P<name>.*)"
+ # end deprecated
+
+ PARAMS = (FS, MO, IP, LP, AD_CIRCLE, AD_RECT, AD_OBROUND, AD_MACRO, AD_POLY, AM, OF, IN, LN)
+ PARAM_STMT = [re.compile(r"%{0}\*%".format(p)) for p in PARAMS]
+
+ COORD_STMT = re.compile((
+ r"(?P<function>{function})?"
+ r"(X(?P<x>{number}))?(Y(?P<y>{number}))?"
+ r"(I(?P<i>{number}))?(J(?P<j>{number}))?"
+ r"(?P<op>{op})?\*".format(number=NUMBER, function=FUNCTION, op=COORD_OP)))
+
+ APERTURE_STMT = re.compile(r"(G54)?D(?P<d>\d+)\*")
+
+ COMMENT_STMT = re.compile(r"G04(?P<comment>[^*]*)(\*)?")
+
+ EOF_STMT = re.compile(r"(?P<eof>M02)\*")
+
+ def __init__(self):
+ self.settings = {}
+ self.statements = []
+
+ def parse(self, filename):
+ fp = open(filename, "r")
+ data = fp.readlines()
+
+ for stmt in self._parse(data):
+ self.statements.append(stmt)
+
+ return GerberFile(self.statements, self.settings, filename)
+
+ def dump_json(self):
+ stmts = {"statements": [stmt.__dict__ for stmt in self.statements]}
+ return json.dumps(stmts)
+
+ def dump_str(self):
+ s = ""
+ for stmt in self.statements:
+ s += str(stmt) + "\n"
+ return s
+
+ def _parse(self, data):
+ oldline = ''
+
+ for i, line in enumerate(data):
+ line = oldline + line.strip()
+
+ # skip empty lines
+ if not len(line):
+ continue
+
+ # deal with multi-line parameters
+ if line.startswith("%") and not line.endswith("%"):
+ oldline = line
+ continue
+
+ did_something = True # make sure we do at least one loop
+ while did_something and len(line) > 0:
+ did_something = False
+
+ # coord
+ (coord, r) = self._match_one(self.COORD_STMT, line)
+ if coord:
+ yield CoordStmt.from_dict(coord, self.settings)
+ line = r
+ did_something = True
+ continue
+
+ # aperture selection
+ (aperture, r) = self._match_one(self.APERTURE_STMT, line)
+ if aperture:
+ yield ApertureStmt(**aperture)
+
+ did_something = True
+ line = r
+ continue
+
+ # comment
+ (comment, r) = self._match_one(self.COMMENT_STMT, line)
+ if comment:
+ yield CommentStmt(comment["comment"])
+ did_something = True
+ line = r
+ continue
+
+ # parameter
+ (param, r) = self._match_one_from_many(self.PARAM_STMT, line)
+ if param:
+ if param["param"] == "FS":
+ stmt = FSParamStmt.from_dict(param)
+ self.settings = {'zero_suppression': stmt.zero_suppression,
+ 'format': stmt.format,
+ 'notation': stmt.notation}
+ yield stmt
+ elif param["param"] == "MO":
+ stmt = MOParamStmt.from_dict(param)
+ self.settings['units'] = stmt.mode
+ yield stmt
+ elif param["param"] == "IP":
+ yield IPParamStmt.from_dict(param)
+ elif param["param"] == "LP":
+ yield LPParamStmt.from_dict(param)
+ elif param["param"] == "AD":
+ yield ADParamStmt.from_dict(param)
+ elif param["param"] == "AM":
+ yield AMParamStmt.from_dict(param)
+ elif param["param"] == "OF":
+ yield OFParamStmt.from_dict(param)
+ elif param["param"] == "IN":
+ yield INParamStmt.from_dict(param)
+ elif param["param"] == "LN":
+ yield LNParamStmtfrom_dict(param)
+ else:
+ yield UnknownStmt(line)
+ did_something = True
+ line = r
+ continue
+
+ # eof
+ (eof, r) = self._match_one(self.EOF_STMT, line)
+ if eof:
+ yield EofStmt()
+ did_something = True
+ line = r
+ continue
+
+ if False:
+ print self.COORD_STMT.pattern
+ print self.APERTURE_STMT.pattern
+ print self.COMMENT_STMT.pattern
+ print self.EOF_STMT.pattern
+ for i in self.PARAM_STMT:
+ print i.pattern
+
+ if line.find('*') > 0:
+ yield UnknownStmt(line)
+ oldline = line
+
+ def _match_one(self, expr, data):
+ match = expr.match(data)
+ if match is None:
+ return ({}, None)
+ else:
+ return (match.groupdict(), data[match.end(0):])
+
+ def _match_one_from_many(self, exprs, data):
+ for expr in exprs:
+ match = expr.match(data)
+ if match:
+ return (match.groupdict(), data[match.end(0):])
+
+ return ({}, None)