Mike Gerber mikegerber

1 netz rosenkohl 1 zwiebel etwas gemüsebrühe

3el erdnussbutter 1-2el sojasoße etwas zitronensaft (nach geschmack) heißes wasser

nicht probiert: zucker

Debian 10
g++-8 8.3.0-6

To reproduce:

git clone https://github.com/OCR-D/olena.git OCR-D-olena
cd OCR-D-olena
./configure --disable-dependency-tracking --enable-scribo
make -j8

The problem

WSL2 uses a random network from the 172.16.0.0/12 RFC1918 private IP address block. And our VPN uses that address block, too, with a route metric of 1 (= most preferred.)

This breaks networking for WSL2. Meh!

The solution

While messing around with the interface/route metric of the VPN network may work around the problem, it also reduces the priority of the VPN. We do not really want this. Additionally, changing the interface metric does not seem to be permanent, so it requires more work when it breaks again.

theme
white

Agenda

💡 Consistency 💡 Less manual work

Part 1: Dealing with versions

	#!/bin/sh
	# Fix mets.xml for easy cases

	xmlstarlet sel -N mets=http://www.loc.gov/METS/ -t -m "//mets:FLocat" -v "@xlink:href" -n mets.xml \| while read f; do
	if echo "$f" \| grep -q "^http://"; then
	continue
	fi

	if [ ! -e $f ]; then
	echo "$f does not exist"

	_jpageviewer_jar=~/opt/jpageviewer/JPageViewer.jar
	if [ -e "$_jpageviewer_jar" ]; then
	jpageviewer() {
	# --resolve-dir defaults to the file's directory
	_jpageviewer_resolve_dir=`dirname "$1"`
	# ... unless a mets.xml file exists one directory down (OCR-D workspace)
	if [ -e "$_jpageviewer_resolve_dir"/../mets.xml ]; then
	_jpageviewer_resolve_dir="$_jpageviewer_resolve_dir"/..
	fi

	from bs4 import BeautifulSoup
	import glob


	METRICS = ["CER", "WER", "WER (order independent)"]
	FILES = glob.glob("*.html")

	print("file;" + ";".join(METRICS))

	import glob
	import json


	FIELDS = ["gt", "ocr", "cer", "wer", "n_characters", "n_words"]
	FILES = glob.glob("*.json")

	print("file;" + ";".join(FIELDS))

	for f in FILES: