summaryrefslogtreecommitdiff
path: root/gerber/excellon_tool.py
diff options
context:
space:
mode:
authorPaulo Henrique Silva <ph.silva@gmail.com>2016-11-16 23:54:09 -0200
committerGitHub <noreply@github.com>2016-11-16 23:54:09 -0200
commit521fe89150c6aaa0ff0954cc8d32e4b6f8009324 (patch)
tree59cef68f1449a5be8d2a9a8ed352e3d9974a3edb /gerber/excellon_tool.py
parent7030eee5e28180732c5cf98aeda4b66657223870 (diff)
parent1e868763a1c4030bedd51553468cbe25f272bef9 (diff)
downloadgerbonara-521fe89150c6aaa0ff0954cc8d32e4b6f8009324.tar.gz
gerbonara-521fe89150c6aaa0ff0954cc8d32e4b6f8009324.tar.bz2
gerbonara-521fe89150c6aaa0ff0954cc8d32e4b6f8009324.zip
Merge pull request #58 from garretfick/merge-curtacircuitos
Epic merge from @garretfick Thanks a lot @garretfick and @hamiltonkibbe.
Diffstat (limited to 'gerber/excellon_tool.py')
-rw-r--r--gerber/excellon_tool.py186
1 files changed, 186 insertions, 0 deletions
diff --git a/gerber/excellon_tool.py b/gerber/excellon_tool.py
new file mode 100644
index 0000000..bd76e54
--- /dev/null
+++ b/gerber/excellon_tool.py
@@ -0,0 +1,186 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+# Copyright 2015 Garret Fick <garret@ficksworkshop.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.
+
+"""
+Excellon Tool Definition File module
+====================
+**Excellon file classes**
+
+This module provides Excellon file classes and parsing utilities
+"""
+
+import re
+try:
+ from cStringIO import StringIO
+except(ImportError):
+ from io import StringIO
+
+from .excellon_statements import ExcellonTool
+
+def loads(data, settings=None):
+ """ Read tool file information and return a map of tools
+ Parameters
+ ----------
+ data : string
+ string containing Excellon Tool Definition file contents
+
+ Returns
+ -------
+ dict tool name: ExcellonTool
+
+ """
+ return ExcellonToolDefinitionParser(settings).parse_raw(data)
+
+class ExcellonToolDefinitionParser(object):
+ """ Excellon File Parser
+
+ Parameters
+ ----------
+ None
+ """
+
+ allegro_tool = re.compile(r'(?P<size>[0-9/.]+)\s+(?P<plated>P|N)\s+T(?P<toolid>[0-9]{2})\s+(?P<xtol>[0-9/.]+)\s+(?P<ytol>[0-9/.]+)')
+ allegro_comment_mils = re.compile('Holesize (?P<toolid>[0-9]{1,2})\. = (?P<size>[0-9/.]+) Tolerance = \+(?P<xtol>[0-9/.]+)/-(?P<ytol>[0-9/.]+) (?P<plated>(PLATED)|(NON_PLATED)|(OPTIONAL)) MILS Quantity = [0-9]+')
+ allegro2_comment_mils = re.compile('T(?P<toolid>[0-9]{1,2}) Holesize (?P<toolid2>[0-9]{1,2})\. = (?P<size>[0-9/.]+) Tolerance = \+(?P<xtol>[0-9/.]+)/-(?P<ytol>[0-9/.]+) (?P<plated>(PLATED)|(NON_PLATED)|(OPTIONAL)) MILS Quantity = [0-9]+')
+ allegro_comment_mm = re.compile('Holesize (?P<toolid>[0-9]{1,2})\. = (?P<size>[0-9/.]+) Tolerance = \+(?P<xtol>[0-9/.]+)/-(?P<ytol>[0-9/.]+) (?P<plated>(PLATED)|(NON_PLATED)|(OPTIONAL)) MM Quantity = [0-9]+')
+ allegro2_comment_mm = re.compile('T(?P<toolid>[0-9]{1,2}) Holesize (?P<toolid2>[0-9]{1,2})\. = (?P<size>[0-9/.]+) Tolerance = \+(?P<xtol>[0-9/.]+)/-(?P<ytol>[0-9/.]+) (?P<plated>(PLATED)|(NON_PLATED)|(OPTIONAL)) MM Quantity = [0-9]+')
+
+ matchers = [
+ (allegro_tool, 'mils'),
+ (allegro_comment_mils, 'mils'),
+ (allegro2_comment_mils, 'mils'),
+ (allegro_comment_mm, 'mm'),
+ (allegro2_comment_mm, 'mm'),
+ ]
+
+ def __init__(self, settings=None):
+ self.tools = {}
+ self.settings = settings
+
+ def parse_raw(self, data):
+ for line in StringIO(data):
+ self._parse(line.strip())
+
+ return self.tools
+
+ def _parse(self, line):
+
+ for matcher in ExcellonToolDefinitionParser.matchers:
+ m = matcher[0].match(line)
+ if m:
+ unit = matcher[1]
+
+ size = float(m.group('size'))
+ platedstr = m.group('plated')
+ toolid = int(m.group('toolid'))
+ xtol = float(m.group('xtol'))
+ ytol = float(m.group('ytol'))
+
+ size = self._convert_length(size, unit)
+ xtol = self._convert_length(xtol, unit)
+ ytol = self._convert_length(ytol, unit)
+
+ if platedstr == 'PLATED':
+ plated = ExcellonTool.PLATED_YES
+ elif platedstr == 'NON_PLATED':
+ plated = ExcellonTool.PLATED_NO
+ elif platedstr == 'OPTIONAL':
+ plated = ExcellonTool.PLATED_OPTIONAL
+ else:
+ plated = ExcellonTool.PLATED_UNKNOWN
+
+ tool = ExcellonTool(None, number=toolid, diameter=size, plated=plated)
+
+ self.tools[tool.number] = tool
+
+ break
+
+ def _convert_length(self, value, unit):
+
+ # Convert the value to mm
+ if unit == 'mils':
+ value /= 39.3700787402
+
+ # Now convert to the settings unit
+ if self.settings.units == 'inch':
+ return value / 25.4
+ else:
+ # Already in mm
+ return value
+
+def loads_rep(data, settings=None):
+ """ Read tool report information generated by PADS and return a map of tools
+ Parameters
+ ----------
+ data : string
+ string containing Excellon Report file contents
+
+ Returns
+ -------
+ dict tool name: ExcellonTool
+
+ """
+ return ExcellonReportParser(settings).parse_raw(data)
+
+class ExcellonReportParser(object):
+
+ # We sometimes get files with different encoding, so we can't actually
+ # match the text - the best we can do it detect the table header
+ header = re.compile(r'====\s+====\s+====\s+====\s+=====\s+===')
+
+ def __init__(self, settings=None):
+ self.tools = {}
+ self.settings = settings
+
+ self.found_header = False
+
+ def parse_raw(self, data):
+ for line in StringIO(data):
+ self._parse(line.strip())
+
+ return self.tools
+
+ def _parse(self, line):
+
+ # skip empty lines and "comments"
+ if not line.strip():
+ return
+
+ if not self.found_header:
+ # Try to find the heaader, since we need that to be sure we understand the contents correctly.
+ if ExcellonReportParser.header.match(line):
+ self.found_header = True
+
+ elif line[0] != '=':
+ # Already found the header, so we know to to map the contents
+ parts = line.split()
+ if len(parts) == 6:
+ toolid = int(parts[0])
+ size = float(parts[1])
+ if parts[2] == 'x':
+ plated = ExcellonTool.PLATED_YES
+ elif parts[2] == '-':
+ plated = ExcellonTool.PLATED_NO
+ else:
+ plated = ExcellonTool.PLATED_UNKNOWN
+ feedrate = int(parts[3])
+ speed = int(parts[4])
+ qty = int(parts[5])
+
+ tool = ExcellonTool(None, number=toolid, diameter=size, plated=plated, feed_rate=feedrate, rpm=speed)
+
+ self.tools[tool.number] = tool \ No newline at end of file