A shell script for collecting Kaitai Struct YAML (.ksy) file paths in GitHub repositories

github-collect-repos-using-ks.sh

#!/bin/sh -f
# SPDX-FileCopyrightText: 2021 Petr Pucil <[email protected]>
#
# SPDX-License-Identifier: MIT

# A shell script for collecting Kaitai Struct YAML (.ksy) file paths in GitHub repositories.
#
# Requires <https://stedolan.github.io/jq> and <https://curl.se> to work.
#
# Inspired by https://github.com/Alhadis/Harvester, but this script:
#   * is much simpler and more portable,
#   * uses the official GitHub REST API returning machine-readable JSON data (instead of
#     requesting HTML pages and parsing their contents),
#   * does not require running a fully-fledged web browser (which also restricts
#     cross-origin HTTP requests so you need to manually navigate to github.com) where you
#     need to log in to your GitHub account,
#   * reduces number of individual HTTP requests (it receives up to 100 results at once -
#     GitHub sends sometimes less but I have no idea why; Harvester can only get 10 results
#     at a time)
#   * saves bandwidth (receives results in JSON, not HTML)
#   * can be conveniently run from command line on any POSIX-compliant shell.

auth_token=<login>:<token> # e.g. octocat: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;
                           # go to <https://github.com/settings/tokens> to get a real one
                           # (no OAuth scopes are needed, public access is enough)
search_query='-org:kaitai-io extension:ksy'
output_file=github-collected-ksy-paths.ndjson
# per_page=100
per_page=$((RANDOM % 61 + 40))
max_page=$(( 1000 / per_page ))
total_count=$(curl \
  -G 'https://api.github.com/search/code' \
  -sS \
  -u "$auth_token" \
  -H "Accept: application/vnd.github.v3+json" \
  --data-urlencode q="${search_query}" \
  --data-urlencode per_page=1 \
  | jq '.total_count')
echo "total count: $total_count"
echo "per page: $per_page"
echo "max page: $max_page"

if [ -f "$output_file" ]; then
  sort -u -o "$output_file" "$output_file"
else
  touch "$output_file"
fi

tmp_f=$(mktemp)
# https://unix.stackexchange.com/a/520041
trap 'rm -vf "$tmp_f"; trap - EXIT; exit' EXIT INT HUP

page=1
num_written_in_run=0
lines_before=$(wc -l < "$output_file")

set --

while true; do
#  if [ "$lines_before" -ge "$total_count" ]; then
#    echo "total count ($total_count) reached, output file has $lines_before lines"
#    break
#  fi

  echo "page $page"
  output=$(curl \
    -G 'https://api.github.com/search/code' \
    -sS \
    -u "$auth_token" \
    -H "Accept: application/vnd.github.v3+json" \
    --data-urlencode q="${search_query}" \
    $@ \
    --data-urlencode per_page="${per_page}" \
    --data-urlencode page="${page}")

  if ! printf %s "$output" \
    | jq --compact-output '.items | map({html_url, repo: .repository.html_url}) | .[]' \
    > "$tmp_f"; then
    echo "jq failed to parse JSON, waiting 65 seconds"
    echo "$output"
    sleep 65
    continue
  fi

  lines_recv=$(wc -l < "$tmp_f")
  if [ "$lines_recv" -ne 0 ]; then
    echo "$lines_recv items received ($per_page requested)"

    LC_ALL=C sort -o "$tmp_f" "$tmp_f"
    LC_ALL=C sort -u -m -o "$output_file" "$output_file" "$tmp_f"
    lines_after=$(wc -l < "$output_file")
    echo "$(( lines_after - lines_before )) items written ($lines_after total)"
    num_written_in_run=$(( num_written_in_run + (lines_after - lines_before) ))
    lines_before=$lines_after
  fi

  if [ $page -ge $max_page ]; then
    page=1
    echo "reached max page, wrapping around to page 1, waiting 50 seconds"
    sleep 50
    if [ $num_written_in_run -lt $max_page ]; then
      if [ $# -lt 2 ]; then
        set -- --data-urlencode order=asc --data-urlencode sort=indexed
      else
        shift 2
      fi
      echo "switching strategy to '$@'"
    fi
    num_written_in_run=0
    per_page=$((RANDOM % 61 + 40))
    max_page=$(( 1000 / per_page ))
    echo "per page: $per_page"
    echo "max page: $max_page"
  else
    page=$((page+1))
    sleep 30
  fi
done

# if [ "$OS" = "Windows_NT" ]; then
#   dos2unix "$output_file" || :
# fi