databases grow indefinitely about avarice HOT 3 CLOSED

Hi Alan,
Thanks for looking at the software. Some of this is due to parts of the software not being upgraded with the rest. Early on when there were fewer indicators and more simple TradeIndicator handling, much of the intent was to have all indicators run constantly for custom strategies to be developed in strategies.py. Some indicators also depend on each other for simpler calculations (e.g. DEMA on EMA, *StochRSI on RSI).
I'm currently working on a set of changes which imperfectly solve this issue. Here are the planned/developed changes:

• Add new objects to each indicator class in indicators.py:
  • IndicatorDepends: list of strings of indicators which are depended on (if applicable).
  • CandleDepends: number of candles which are needed to get results with current configuration.
• Add a new argument to storage.writelist for how many objects should be held in the each list.
• Only use TradeIndicators and IndicatorDepends for running indicators.
• Support printing number of required candles before getting enough info/trading (uses CandleDepends).
• Using per-indicator shelve files with all genconfig indicator class objects included in the name. This fixes a bug if the software is resumed with changed configuration. This also avoids the need for a queue while again using concurrent futures for running indicators.
• Only store the max CandleDepends number of candles in sqlite. While generally only the last candle is needed once the indicator is producing output, this is a far simpler approach and still keeps size down tremendously.

from avarice.

I finally got a chance to work on these changes. I'm currently testing them in https://github.com/Galts-Gulch/avarice/commits/indicator-rewrite

from avarice.

That branch has been merged, and the release has been tagged. This issue is now mostly fixed.

from avarice.