QuantRocket

Performance improvement when using Reuters estimates with large datasets

release-notes

#1

Anyone working with Reuters estimates using get_reuters_estimates_reindexed_like is encouraged to update to quantrocket/jupyter:1.2.1 and quantrocket/moonshot:1.2.1 to get an updated version of the function.

Problem: calling get_reuters_estimates_reindexed_like on a large universe of securities (e.g. all US stocks) was unusably slow.

Reason: Unlike Reuters financials, which are indexed by date, Reuters estimates are indexed by datetime. Due to Reuters typically providing a different datetime for each security, the reindexing process was very slow for large universes.

Solution: Datetimes are now cast to dates before reindexing. This greatly improves performance. Note that downcasting from datetimes to dates doesn’t alter the resulting DataFrame because it was already being returned at daily (not intraday) resolution.

For additional details and instructions on updating, see the Release Notes.


Fundamental codes missing for Reuters Estimates