David Wilson
David Wilson
repro: echo foo | doas cat This breaks piping something to a command via doas, and breaks tools (dw/mitogen) expecting it to behave like the BSD version
Consider: - Track the maximum output size of any struct on a per-StructType basis, and use that as StringBuffer's size hint. --- By abusing TLS, could change the interface to...
[final design: http://pythonsweetness.tumblr.com/post/73248785592/acid-batch-record-format ] 1. "Pure keys" mode: when a collection's key is based entirely on the record value (e.g. log line timestamp) or a common prefix, batches need only...
This depends at least on finalizing a new batch key representation (#30) and writing a Python iterator for it, and partially on picking a final solution for #23.
The goal is to exploit database engines like LMDB (via py-lmdb) by transparently extending their zero-copy behaviour up into user code. Since LMDB never copies during reads, range scans on...
Now that transaction events are mostly implemented (#78), various aggregate functions could easily be provided, e.g. at least COUNT, SUM, and AVG. But using indices, it might be possible to...
Consider: Introduce new Range type composed of something like {store, prefix, mapper} Range has: ``` prefix(tup) -> subrange range(lo, hi, include=True) -> subrange get(tup, reverse=False) -> fetch prefix+tuple find(tup, reverse=False)...
Per ticket #39 and https://github.com/dw/acid/commit/82d00fccb0f0fa117e9b7fd2bac4aea9c61f4903 , move events (nee "triggers") out of meta.py and support them directly in core.py. Additionally add support for `after_replace`, which can be supported using the...
Currently a request context is simply an object that manages a reference to the active transaction for the duration of some conceptual "request". The default request context implementation uses TLS,...
There is still value in secondary indexing of hashed collections, and it would be possible to support with a small addition to the Engine interface, and the completion of #70