import subprocess from pathlib import Path import tempfile import os from functools import total_ordering import shutil import bs4 from contextlib import contextmanager import numpy as np from PIL import Image @total_ordering class ImageDifference: def __init__(self, value, histogram): self.value = value self.histogram = histogram def __float__(self): return float(self.value) def __eq__(self, other): return float(self) == float(other) def __lt__(self, other): return float(self) < float(other) def __str__(self): return str(float(self)) @total_ordering class Histogram: def __init__(self, value, size): self.value, self.size = value, size def __eq__(self, other): other = np.array(other) other[other == None] = self.value[other == None] return (self.value == other).all() def __lt__(self, other): other = np.array(other) other[other == None] = self.value[other == None] return (self.value <= other).all() def __getitem__(self, index): return self.value[index] def __str__(self): return f'{list(self.value)} size={self.size}' def run_cargo_cmd(cmd, args, **kwargs): if cmd.upper() in os.environ: return subprocess.run([os.environ[cmd.upper()], *args], **kwargs) try: return subprocess.run([cmd, *args], **kwargs) except FileNotFoundError: return subprocess.run([str(Path.home() / '.cargo' / 'bin' / cmd), *args], **kwargs) def svg_to_png(in_svg, out_png): run_cargo_cmd('resvg', ['--dpi', '100', in_svg, out_png], check=True, stdout=subprocess.DEVNULL) def gbr_to_svg(in_gbr, out_svg, origin=(0, 0), size=(6, 6)): x, y = origin w, h = size cmd = ['gerbv', '-x', 'svg', '--border=0', f'--origin={x:.6f}x{y:.6f}', f'--window_inch={w:.6f}x{h:.6f}', '--foreground=#ffffff', '-o', str(out_svg), str(in_gbr)] subprocess.run(cmd, check=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) @contextmanager def svg_soup(filename): with open(filename, 'r') as f: soup = bs4.BeautifulSoup(f.read(), 'xml') yield soup with open(filename, 'w') as f: f.write(str(soup)) def cleanup_clips(soup): for group in soup.find_all('g'): # gerbv uses Cairo's SVG canvas. Cairo's SVG canvas is kind of broken. It has no support for unit # handling at all, which means the output files just end up being in pixels at 72 dpi. Further, it # seems gerbv's aperture macro rendering interacts poorly with Cairo's SVG export. gerbv renders # aperture macros into a new surface, which for some reason gets clipped by Cairo to the given # canvas size. This is just wrong, so we just nuke the clip path from these SVG groups here. # # Apart from being graphically broken, this additionally causes very bad rendering performance. del group['clip-path'] # remove broken clip def gerber_difference(reference, actual, diff_out=None, svg_transform=None, size=(10,10)): with tempfile.NamedTemporaryFile(suffix='.svg') as act_svg,\ tempfile.NamedTemporaryFile(suffix='.svg') as ref_svg: gbr_to_svg(reference, ref_svg.name, size=size) gbr_to_svg(actual, act_svg.name, size=size) with svg_soup(ref_svg.name) as soup: if svg_transform is not None: soup.find('g', attrs={'id': 'surface1'})['transform'] = svg_transform cleanup_clips(soup) with svg_soup(act_svg.name) as soup: cleanup_clips(soup) # FIXME DEBUG shutil.copyfile(act_svg.name, '/tmp/test-act.svg') shutil.copyfile(ref_svg.name, '/tmp/test-ref.svg') return svg_difference(ref_svg.name, act_svg.name, diff_out=diff_out) def svg_difference(reference, actual, diff_out=None): with tempfile.NamedTemporaryFile(suffix='-ref.png') as ref_png,\ tempfile.NamedTemporaryFile(suffix='-act.png') as act_png: svg_to_png(reference, ref_png.name) svg_to_png(actual, act_png.name) return image_difference(ref_png.name, act_png.name, diff_out=diff_out) def image_difference(reference, actual, diff_out=None): ref = np.array(Image.open(reference)).astype(float) out = np.array(Image.open(actual)).astype(float) ref, out = ref.mean(axis=2), out.mean(axis=2) # convert to grayscale # FIXME blur images here before comparison to mitigate aliasing issue delta = np.abs(out - ref).astype(float) / 255 if diff_out: Image.fromarray((delta*255).astype(np.uint8), mode='L').save(diff_out) hist, _bins = np.histogram(delta, bins=10, range=(0, 1)) return (ImageDifference(delta.mean(), hist), ImageDifference(delta.max(), hist), Histogram(hist, out.size))