Skip to content

0.9.0

Choose a tag to compare

View all tags
@greenape greenape released this 01 Oct 10:24
· 8067 commits to master since this release
0.9.0
870fd31

Added

  • FlowAuth now makes version information available at /version and displays it in the web ui. #835
  • FlowETL now comes with a deployment example (in flowetl/deployment_example/). #1126
  • FlowETL now allows to run supplementary post-ETL queries. #989
  • Random sampling is now exposed via the API, for all non-aggregated query kinds. #1007
  • New aggregate added to FlowMachine - HistogramAggregation, which constructs histograms over the results of other queries. #1075
  • New IntereventInterval query class - returns stats over the gap between events as a time interval.
  • Added submodule flowmachine.core.dependency_graph, which contains functions related to creating or using query dependency graphs (previously these were in utils.py).
  • New config option sql_find_available_dates in FlowETL to provide SQL code to determine the available dates. #1295

Changed

  • FlowDB is now based on PostgreSQL 11.5 and PostGIS 2.5.3
  • When running queries through FlowAPI, the query's dependencies will also be cached by default. This behaviour can be switched off by setting FLOWMACHINE_SERVER_DISABLE_DEPENDENCY_CACHING=true. #1152
  • NewSubscribers now takes a pair of UniqueSubscribers queries instead of the arguments to them
  • Flowmachine's default random sampling method is now random_ids rather than the non-reproducible system_rows. #1263
  • IntereventPeriod now returns stats over the gap between events in fractional time units, instead of time intervals. #1265
  • Attempting to store a query that does not have a standard table name (e.g. EventTableSubset or unseeded random sample) will now raise an UnstorableQueryError instead of ValueError.
  • In the FlowETL deployment example, the external ingestion database is now set up separately from the FlowKit components and connected to FlowDB via a docker overlay network. #1276
  • The md5 attribute of the Query class has been renamed to query_id #1288.
  • DistanceMatrix no longer returns duplicate rows for the lon-lat spatial unit.
  • Previously, Displacement defaulted to returning NaN for subscribers who have a location in the reference location but were not seen in the time period for the displacement query. These subscribers are no longer returned unless the return_subscribers_not_seen argument is set to True.
  • PopulationWeightedOpportunities is now available under flowmachine.features.location, instead of flowmachine.models
  • PopulationWeightedOpportunities no longer supports erroring with incomplete per-location departure rate vectors and will instead omit any locations not included from the results
  • PopulationWeightedOpportunities no longer requires use of the run() method

Fixed

  • Quickstart will no longer fail if it has been run previously with a different FlowDB data size and not explicitly shut down. #900

Removed

  • Flowmachine's subscriber_locations_cluster function has been removed - use HartiganCluster or MeaningfulLocations directly.
  • FlowAPI no longer supports the non-reproducible random sampling method system_rows. #1263
Assets 2
Loading