Merge pull request #8831 from bhandras/sql-invoice-migration

invoices: migrate KV invoices to native SQL for users of KV SQL backends

scripts/gen_sqlc_docker.sh

@@ -2,12 +2,40 @@
 
 set -e
 
+# restore_files is a function to restore original schema files.
+restore_files() {
+	echo "Restoring SQLite bigint patch..."
+	for file in sqldb/sqlc/migrations/*.up.sql.bak; do
+		mv "$file" "${file%.bak}"
+	done
+}
+
+
+# Set trap to call restore_files on script exit. This makes sure the old files
+# are always restored.
+trap restore_files EXIT
+
 # Directory of the script file, independent of where it's called from.
 DIR="$(cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd)"
 # Use the user's cache directories
 GOCACHE=`go env GOCACHE`
 GOMODCACHE=`go env GOMODCACHE`
 
+# SQLite doesn't support "BIGINT PRIMARY KEY" for auto-incrementing primary
+# keys, only "INTEGER PRIMARY KEY". Internally it uses 64-bit integers for
+# numbers anyway, independent of the column type. So we can just use
+# "INTEGER PRIMARY KEY" and it will work the same under the hood, giving us
+# auto incrementing 64-bit integers.
+# _BUT_, sqlc will generate Go code with int32 if we use "INTEGER PRIMARY KEY",
+# even though we want int64. So before we run sqlc, we need to patch the
+# source schema SQL files to use "BIGINT PRIMARY KEY" instead of "INTEGER
+# PRIMARY KEY".
+echo "Applying SQLite bigint patch..."
+for file in sqldb/sqlc/migrations/*.up.sql; do
+	echo "Patching $file"
+	sed -i.bak -E 's/INTEGER PRIMARY KEY/BIGINT PRIMARY KEY/g' "$file"
+done
+
 echo "Generating sql models and queries in go..."
 
 docker run \