Historical package download data is used to display download charts on project pages as well as for the calculation of trending projects.

To keep the usage of database resources at bay we persist only weekly stats, the cutoff day being Sunday.

The download data is being synced from the package registry API throughout each day, so the numbers do not have an equal cutoff time - therefore they will always be slightly skewed. For example one project might get it's stats updated early in the day, another towards the end of the day, meaning the number of downloads for the latter has effectively accumulated downloads from another day. As "incorrect" as this may be, for the purposes of illustrating historical trends in package downloads we consider this "good enough" ;)

How to read the charts

The charts show total accumulated downloads as well as the absolute change over the trailing month. Keep in mind that install counts include automated installs from CI systems and deployments, so they indicate relative popularity between libraries rather than an exact number of human users.

Because the dataset is only sampled once per week, short-lived spikes may not be visible in the charts.