Currently, Runner subclasses keep track of whatever output files they wrote in the attribute output_files, which is stored in the JSON payload of the DB. This is necessary because of awkwardness involving relative paths and finding the directory a taxi is working in. This scheme is fragile and a little graceless. Instead, after completing a task, have the taxi store the working directory a task was executed in in the dispatch DB. This should probably be a new column in the task table, but could be part of the JSON payload at first.
Currently, Runner subclasses keep track of whatever output files they wrote in the attribute
output_files, which is stored in the JSON payload of the DB. This is necessary because of awkwardness involving relative paths and finding the directory a taxi is working in. This scheme is fragile and a little graceless. Instead, after completing a task, have the taxi store the working directory a task was executed in in the dispatch DB. This should probably be a new column in the task table, but could be part of the JSON payload at first.