#!/usr/bin/env bash
set -euo pipefail

# Usage: ./fetch_stats_from_ci.sh
#
# This script is for fetching historic test stats from GitHub Actions CI.
#
# Requires gh with credentials.
#
# https://github.com/cli/cli/blob/trunk/pkg/cmd/run/view/view.go#L434

dir="$(dirname "$0")"/ci-stats
mkdir -p "${dir}"

# Disable gh run view logs, it's unreliable.
USE_GH_RUN_VIEW_LOGS=0

pushd "${dir}" >/dev/null

# Stats step name, used for filtering log.
job_step_name="Print test stats"

if [[ ! -f list-ci.yaml.json ]]; then
	gh run list -w ci.yaml -L 1000 --json conclusion,createdAt,databaseId,displayTitle,event,headBranch,headSha,name,number,startedAt,status,updatedAt,url,workflowDatabaseId,workflowName \
		>list-ci.yaml.json || {
		rm -f list-ci.yaml.json
		exit 1
	}
fi

runs="$(
	jq -r '.[] | select(.status == "completed") | select(.conclusion == "success" or .conclusion == "failure") | [.databaseId, .event, .displayTitle, .headBranch, .headSha, .url] | @tsv' \
		<list-ci.yaml.json
)"

while read -r run; do
	mapfile -d $'\t' -t parts <<<"${run}"
	parts[-1]="${parts[-1]%$'\n'}"

	database_id="${parts[0]}"
	event="${parts[1]}"
	display_title="${parts[2]}"
	head_branch="${parts[3]}"
	head_sha="${parts[4]}"
	run_url="${parts[5]}"

	# Check if this run predates the stats PR, if yes, skip it:
	# https://github.com/coder/coder/issues/6676
	if [[ ${database_id} -le 4595490577 ]]; then
		echo "Skipping ${database_id} (${display_title}), too old..."
		continue
	fi

	run_jobs_file=run-"${database_id}"-"${event}"-jobs.json
	if [[ ! -f "${run_jobs_file}" ]]; then
		echo "Fetching jobs for run: ${display_title} (${database_id}, ${event}, ${head_branch})"
		gh run view "${database_id}" --json jobs >"${run_jobs_file}" || {
			rm -f "${run_jobs_file}"
			exit 1
		}
	fi

	jobs="$(
		jq -r '.jobs[] | select(.name | startswith("test-go")) | select(.status == "completed") | select(.conclusion == "success" or .conclusion == "failure") | [.databaseId, .startedAt, .completedAt, .name, .url] | @tsv' \
			<"${run_jobs_file}"
	)"

	while read -r job; do
		mapfile -d $'\t' -t parts <<<"${job}"
		parts[-1]="${parts[-1]%$'\n'}"

		job_database_id="${parts[0]}"
		job_started_at="${parts[1]}"
		job_completed_at="${parts[2]}"
		job_name="${parts[3]}"
		job_url="${parts[4]}"

		job_log=run-"${database_id}"-job-"${job_database_id}"-"${job_name}".log
		if [[ ! -f "${job_log}" ]]; then
			echo "Fetching log for: ${job_name} (${job_database_id}, ${job_url})"

			if [[ ${USE_GH_RUN_VIEW_LOGS} -eq 0 ]]; then
				# Since gh run view is unreliable, we will fetch the logs via API
				# instead, however, unfortunately the API does not provide the job
				# name in the log output.
				#
				# TODO(mafredri): This would be more reliably fetched from the following URL:
				# https://github.com/coder/coder/commit/${head_sha}/checks/${job_database_id}/logs/${job_step_number}
				# but it requires browser-level authentication(?).
				#
				# Example output:
				#
				#   2023-04-14T05:43:34.4763012Z ##[group]Run # Artifacts are not available after rerunning a job,
				#   2023-04-14T05:43:34.4763385Z # Artifacts are not available after rerunning a job,
				#   2023-04-14T05:43:34.4763815Z # so we need to print the test stats to the log.
				#   2023-04-14T05:43:34.4764149Z go run ./scripts/ci-report/main.go gotests.json | tee gotests_stats.json
				#   2023-04-14T05:43:34.4809056Z shell: /usr/bin/bash -e {0}
				#   2023-04-14T05:43:34.4809308Z ##[endgroup]
				#   2023-04-14T05:43:35.5934784Z {
				#   2023-04-14T05:43:35.5935419Z   "packages": [
				#   2023-04-14T05:43:35.5936020Z     {
				#   2023-04-14T05:43:35.5936585Z       "name": "agent",
				#   2023-04-14T05:43:35.5937105Z       "time": 17.044
				#   2023-04-14T05:43:35.5937631Z     },
				gh api "/repos/coder/coder/actions/jobs/${job_database_id}/logs" >"${job_log}" || {
					# Sometimes gh fails to extract ZIP, etc. :'(
					rm -f "${job_log}"
					echo "Failed to fetch log for: ${job_name} (${job_database_id}, ${job_url}), skipping..."
					continue
				}

				# Elaborate loop for finding the starting point for $job_step_name.
				# We check for the first occurrence of "##[group]" which contains
				# the go run command and then continue until we find the next
				# "##[group]". We then print everything in between.
				log_buffer=()
				found_step=0
				while read -r line; do
					if [[ ${found_step} -eq 1 ]] && [[ ${#log_buffer[@]} -eq 0 ]]; then
						if [[ ${line} == *"##[group]"* ]]; then
							break
						fi
						# Mimic output from gh run view.
						echo "${job_name}"$'\t'"${job_step_name}"$'\t'"${line}"
					fi
					if [[ ${found_step} -eq 0 ]] && [[ ${#log_buffer[@]} -eq 0 ]] && [[ ${line} != *"##[group]"* ]]; then
						continue
					fi
					if [[ ${line} == *"##[group]"* ]]; then
						log_buffer=("${line}")
						continue
					fi
					if [[ ${#log_buffer[@]} -gt 0 ]]; then
						log_buffer+=("${line}")
					fi
					if [[ ${line} == *"##[endgroup]"* ]]; then
						if [[ ${found_step} -eq 1 ]]; then
							for bufline in "${log_buffer[@]}"; do
								# Mimic output from gh run view.
								echo "${job_name}"$'\t'"${job_step_name}"$'\t'"${bufline}"
							done
						fi
						log_buffer=()
						continue
					fi
					# If line contains go run ./scripts/ci-report/main.go gotests.json
					if [[ ${line} == *"go run ./scripts/ci-report/main.go"* ]]; then
						found_step=1
					fi
				done <"${job_log}" >"${job_log}.parsed"
				mv "${job_log}.parsed" "${job_log}"
			else
				# Example log (partial).
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:18.4063489Z ##[group]Run # Artifacts are not available after rerunning a job,
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:18.4063872Z # Artifacts are not available after rerunning a job,
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:18.4064188Z # so we need to print the test stats to the log.
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:18.4064642Z go run ./scripts/ci-report/main.go gotests.json | tee gotests_stats.json
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:18.4110112Z shell: /usr/bin/bash -e {0}
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:18.4110364Z ##[endgroup]
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:19.3440469Z {
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:19.3441078Z   "packages": [
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:19.3441448Z     {
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:19.3442927Z       "name": "agent",
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:19.3443311Z       "time": 17.538
				#   test-go (ubuntu-latest)	Print test stats	2023-04-11T03:02:19.3444048Z     },
				#   ...
				gh run view --job "${job_database_id}" --log >"${job_log}" || {
					# Sometimes gh fails to extract ZIP, etc. :'(
					rm -f "${job_log}"
					echo "Failed to fetch log for: ${job_name} (${job_database_id}, ${job_url}), skipping..."
					continue
				}
			fi
		fi

		log_lines="$(wc -l "${job_log}" | awk '{print $1}')"
		if [[ ${log_lines} -lt 7 ]]; then
			# Sanity check in case something went wrong, the ##[group]
			# and ##[endgroup] header is 6 lines and start of JSON ("{")
			# makes the 7th.
			rm -f "${job_log}"
			echo "Log is empty for: ${job_name} (${job_database_id}, ${job_url}), skipping..."
			continue
		fi

		if ! job_stats="$(
			# Extract the stats job output (JSON) from the job log,
			# discarding the timestamp and non-JSON header.
			#
			# Example variable values:
			#   job_name="test-go (ubuntu-latest)"
			#   job_step_name="Print test stats"
			grep "${job_name}.*${job_step_name}" "${job_log}" |
				sed -E 's/.*[0-9-]{10}T[0-9:]{8}\.[0-9]*Z //' |
				grep -E "^[{}\ ].*"
		)"; then
			echo "Failed to find stats in job log: ${job_name} (${job_database_id}, ${job_url}), skipping..."
			continue
		fi

		if ! jq -e . >/dev/null 2>&1 <<<"${job_stats}"; then
			# Sometimes actions logs are partial when fetched via CLI :'(
			echo "Failed to parse stats for: ${job_name} (${job_database_id}, ${job_url}), skipping..."
			continue
		fi

		job_stats_file=run-"${database_id}"-job-"${job_database_id}"-"${job_name}"-stats.json
		if [[ -f "${job_stats_file}" ]]; then
			continue
		fi
		jq \
			--argjson run_id "${database_id}" \
			--arg run_url "${run_url}" \
			--arg event "${event}" \
			--arg branch "${head_branch}" \
			--arg sha "${head_sha}" \
			--arg started_at "${job_started_at}" \
			--arg completed_at "${job_completed_at}" \
			--arg display_title "${display_title}" \
			--argjson job_id "${job_database_id}" \
			--arg job "${job_name}" \
			--arg job_url "${job_url}" \
			'{run_id: $run_id, run_url: $run_url, event: $event, branch: $branch, sha: $sha, started_at: $started_at, completed_at: $completed_at, display_title: $display_title, job_id: $job_id, job: $job, job_url: $job_url, stats: .}' \
			<<<"${job_stats}" \
			>"${job_stats_file}" || {
			echo "Failed to write stats for: ${job_name} (${job_database_id}, ${job_url}), skipping..."
			rm -f "${job_stats_file}"
			exit 1
		}
	done <<<"${jobs}"
done <<<"${runs}"