# Toolchain for the Ross-1988 re-typeset project (matches the environment the
# first two sessions ran in: Ubuntu 24.04, TeX Live 2023, tesseract 5.3, poppler).
FROM ubuntu:24.04
ENV DEBIAN_FRONTEND=noninteractive LANG=C.UTF-8
RUN apt-get update && apt-get install -y --no-install-recommends \
      texlive-xetex texlive-latex-base texlive-latex-recommended texlive-latex-extra \
      texlive-fonts-recommended texlive-fonts-extra \
      fonts-liberation fonts-freefont-ttf fontconfig \
      poppler-utils tesseract-ocr tesseract-ocr-eng \
      python3 python3-pip python3-pil python3-numpy \
      make curl ca-certificates \
      qpdf ghostscript img2pdf \
      git automake autoconf libtool g++ pkg-config libleptonica-dev zlib1g-dev \
 && rm -rf /var/lib/apt/lists/* \
 && pip3 install --no-cache-dir --break-system-packages pdfplumber fonttools \
 && fc-cache -f

# jbig2enc: the encoder the ProQuest scan itself uses for its bilevel page
# layer, and the reason a 2556x3305 page fits in 4900 bytes there. It builds a
# dictionary of repeated letterforms and stores each shape once, which general
# compressors cannot do -- measured on this scan, 0.019 bits/pixel against
# 0.084 for CCITT G4 and 0.159 for 1-bit PNG. Not packaged for Ubuntu 24.04, so
# it is built from source.
RUN git clone --depth 1 https://github.com/agl/jbig2enc /tmp/jbig2enc \
 && cd /tmp/jbig2enc \
 && ./autogen.sh && ./configure && make -j"$(nproc)" && make install \
 && ldconfig \
 && rm -rf /tmp/jbig2enc
WORKDIR /work
# The project directory is bind-mounted at /work (see compose.yaml); nothing is copied in.
CMD ["bash"]