#!/usr/bin/env bash

build-download-zip

#

WHAT (for a CEO): builds the single "download everything" file offered on the

source browser -- one zip containing both the public website AND the source

code that generates it, so anyone can take the whole thing home and rebuild it.

#

PRIVACY: the source files are exactly git ls-files (which obeys .gitignore,

so the raw private inputs -- the poem corpus, LLM transcripts, archives --

never appear) filtered to the SAME allowlist the source browser publishes:

the src/libs/scripts/issues/docs/notes/demos directories plus code/doc files at

the root. If you change that allowlist in generate-source-browser.lua, change

it here too (the two must agree, or the download shows more/less than the

browser). The rendered site (output/, itself gitignored) is already the public

deployable, so it ships whole.

#

HOW: assembles the archive in /tmp (RAM, and outside output/ so it can't zip

itself), then moves the finished file into output/<NAME>.zip, served at

/similar-different/<NAME>.zip.

#

Usage: scripts/build-download-zip [DIR]

set -euo pipefail

{{{ configuration

${DIR} is the project root; hard-coded for the normal case, overridable as the

first argument so the script runs from anywhere.

DIR="${1:-/mnt/mtwo/programming/ai-stuff/neocities-modernization}"
NAME="similar-different"

Build in RAM, outside output/, so the archive never tries to contain itself.

ZIP_TMP="/tmp/neocities-modernization/${NAME}.zip"
FINAL="${DIR}/output/${NAME}.zip"

}}}

{{{ preflight

if [ ! -d "$DIR/output" ]; then
echo "[build-download-zip] ERROR: $DIR/output not found -- generate the site first." >&2
exit 1
fi
mkdir -p "$(dirname "$ZIP_TMP")"
rm -f "$ZIP_TMP"

}}}

{{{ source files (allowlisted, git-tracked -> respects .gitignore)

A path WITH a slash is kept only if its top directory is allowlisted; a root

file is kept only if its extension is a code/doc type. This mirrors

generate-source-browser.lua's INCLUDE_DIRS / ROOT_FILE_EXTS exactly.

SOURCE_LIST="$(cd "$DIR" && git ls-files | awk -F/ '
NF > 1 { if ($1 ~ /^(src|libs|scripts|issues|docs|notes|demos)$/) print; next }
{ if ($0 ~ /\.(lua|sh|md|c|h|json|css|html|txt)$/) print }
')"
printf '%s\n' "$SOURCE_LIST" | (cd "$DIR" && zip -q "$ZIP_TMP" -@)

}}}

{{{ LICENSE (added explicitly: it may not be committed yet, so git ls-files

would miss it, but it must travel with the bundle it governs)

if [ -f "$DIR/LICENSE" ]; then
(cd "$DIR" && zip -q "$ZIP_TMP" LICENSE)
fi

}}}

{{{ readme.txt -- orient whoever unzips this (how to open it, what's inside)

README_TMP="$(dirname "$ZIP_TMP")/readme.txt"
cat > "$README_TMP" <<'EOF'
similar-different -- the poetry site, plus the source that builds it.

HOW TO OPEN IT
The pages use document-relative links, so just open output/wordcloud.html
(or output/index.html) in any web browser, from wherever you unzipped this, on
any computer. Images, navigation, the source browser -- all resolve. No web
server needed.

WHAT'S INSIDE
output/ the rendered website (what you'd see online), with relative links
src/, libs/, scripts/ the Lua that generates it; run.sh drives the pipeline
issues/, docs/, notes/ the design record -- how and why it was built
LICENSE what you may do with all of it (everything; just don't be cruel)

ABOUT THE LINKS
The pages use document-relative links (../media/x.png, ../wordcloud.html), so
they resolve from anywhere with no conversion -- on your disk or re-hosted. The
generators emit this form directly; there is no build step that "fixes" paths.
To regenerate the whole site from this source, run run.sh.
EOF
(cd "$(dirname "$README_TMP")" && zip -q "$ZIP_TMP" readme.txt)
rm -f "$README_TMP"

}}}

{{{ the rendered site (output/ -> stored under output/ inside the zip)

Exclude any previous build's archive sitting in output/ -- otherwise each run's

2.6G zip would swallow the last run's, doubling forever.

(cd "$DIR" && zip -rq "$ZIP_TMP" output -x "output/${NAME}.zip")

}}}

{{{ publish: move the finished archive into output/ for serving

mv -f "$ZIP_TMP" "$FINAL"
SIZE="$(du -h "$FINAL" | cut -f1)"
COUNT="$(printf '%s\n' "$SOURCE_LIST" | grep -c . || true)"
echo "[build-download-zip] built $FINAL ($SIZE): $COUNT source files + the rendered site"

}}}