Open bartread opened 8 years ago
I think caching the TypePropertyMaps as well might be a good shout, based on what ANTS Performance Profiler is showing me.
It also seems as though the most expensive part of the query generation is creating the SELECT list.
Generating auto-queries can be expensive since reflection is used to walk over the types and generate SQL. The SQL itself can become quite large so there's a lot of appending to StringBuilders happening, which will obviously cause allocations and copying.
It would be better to cache auto-queries based on criteria such as:
You don't need to cache anything for calls to CustomQuery because Dapper will cache the queries itself. With AutoQueries this is just a way of avoiding large amounts of SQL generation for frequently repeated queries, which should noticeably improve performance under load.