The Anaconda project publishes summarized download counts of conda packages distributed via the Anaconda Distribution, and the conda-forge and bioconda channels (https://www.anaconda.com/announcing-public-anaconda-package-download-data/). The data are available in the Parquet file format on Amazon S3 and can be parsed relatively easily in Python (full details at https://github.com/ContinuumIO/anaconda-package-data).
However reading these data this is non-trivial in R. This repository provides TSV and RDS tables of monthly download counts for all packages, aggregated across all platforms and versions. This summarisation level was chosen to match the download statistics reported by the Bioconductor project (http://bioconductor.org/packages/stats/) and to that end we also make available count tables containing only Bioconductor packages distributed via bioconda.
Tab-separated and serialized R-object files for the two count tables can be found in the tsv and rdata folders respectively. These are updated monthly with some lag behind the release of the data from Anaconda. The TSV files are UTF-8 encoded text files and the RDS file contain the data as a tibble.
Summarised counts for all packages distributed via Anaconda can be found in all_counts.tsv and all_counts.rds. Both are four column tables, with column names preserved from the Anaconda source files e.g.
all_counts %>% filter(pkg_name == "bioconductor-deseq2")
# A tibble: 29 x 4
pkg_name year month counts
<chr> <chr> <chr> <dbl>
1 bioconductor-deseq2 2017 Mar 88
2 bioconductor-deseq2 2017 Apr 709
3 bioconductor-deseq2 2017 May 765
4 bioconductor-deseq2 2017 Jun 876
# … with 25 more rowsThe bioc_counts.tsv and bioc_counts.rds files contain a subset of the download data relating only to Bioconductor packages. In these tables the package names have been transformed from their bioconda format (all lower case, prefixed with bioconductor) to how they appear in the Bioconductor repository e.g. bioconductor-deseq2 🠪 DESeq2. The column names have also been changed to match the tables produced by Bioconductor e.g.
bioc_counts %>% filter(Package == "DESeq2")
# A tibble: 29 x 4
Package Year Month Nb_of_downloads
<chr> <chr> <chr> <dbl>
1 DESeq2 2017 Mar 88
2 DESeq2 2017 Apr 709
3 DESeq2 2017 May 765
4 DESeq2 2017 Jun 876
# … with 25 more rowsAccess to the Bioconductor count table is also provided via the anacondaDownloadStats() function in BiocPkgTools. You will need version 1.9.2 or greater of the package to have access to the function. The table of counts is formatted by the function to be consistent with other data retrieval functions within the package.
BiocPkgTools::anacondaDownloadStats() %>% filter(Year == 2019, Month == "Jan")
# A tibble: 1,342 x 7
Package Year Month Nb_of_distinct_IPs Nb_of_downloads repo Date
<chr> <chr> <chr> <int> <dbl> <chr> <date>
1 a4 2019 Jan NA 1 Anaconda 2019-01-01
2 a4Base 2019 Jan NA 147 Anaconda 2019-01-01
3 a4Classif 2019 Jan NA 3 Anaconda 2019-01-01
4 a4Core 2019 Jan NA 226 Anaconda 2019-01-01
5 a4Preproc 2019 Jan NA 220 Anaconda 2019-01-01
# … with 1,337 more rowsScripts for producing the count tables can be found in the R folder:
download_functions.R: Contains functions to download and process the parquet files containing the package download counts.update_tables.R: For all months from January 2017 to today, downloads and collates the daily count data into the complete tables. Intended to be run on a monthly basis via GitHub actions.