Download and Aggregate Data from Public Hire Bicycle Systems

Download and aggregate data from all public hire bicycle systems which provide open data, currently including 'Santander' Cycles in London, U.K.; from the U.S.A., 'Ford GoBike' in San Francisco CA, 'citibike' in New York City NY, 'Divvy' in Chicago IL, 'Capital Bikeshare' in Washington DC, 'Hubway' in Boston MA, 'Metro' in Los Angeles LA, 'Indego' in Philadelphia PA, and 'Nice Ride' in Minnesota; 'Bixi' from Montreal, Canada; and 'mibici' from Guadalajara, Mexico.

BuildStatus Buildstatus codecov Project Status:Active CRAN_Status_Badge CRANDownloads

The bikedata package aims to enable ready importing of historical trip data from all public bicycle hire systems which provide data, and will be expanded on an ongoing basis as more systems publish open data. Cities and names of associated public bicycle systems currently included, along with numbers of bikes and of docking stations (from wikipedia), are

City Hire Bicycle System Number of Bicycles Number of Docking Stations
London, U.K. Santander Cycles 13,600 839
San Francisco Bay Area, U.S.A. Ford GoBike 7,000 540
New York City NY, U.S.A. citibike 7,000 458
Chicago IL, U.S.A. Divvy 5,837 576
Montreal, Canada Bixi 5,220 452
Washingon DC, U.S.A. Capital BikeShare 4,457 406
Guadalajara, Mexico mibici 2,116 242
Minneapolis/St Paul MN, U.S.A. Nice Ride 1,833 171
Boston MA, U.S.A. Hubway 1,461 158
Philadelphia PA, U.S.A. Indego 1,000 105
Los Angeles CA, U.S.A. Metro 1,000 65

These data include the places and times at which all trips start and end. Some systems provide additional demographic data including years of birth and genders of cyclists. The list of cities may be obtained with the bike_cities() functions, and details of which include demographic data with bike_demographic_data().

The following provides a brief overview of package functionality. For more detail, see the vignette.

1 Installation

Currently a development version only which can be installed with the following command,


and then loaded the usual way

library (bikedata)

2 Usage

Data may downloaded for a particular city and stored in an SQLite3 database with the simple command,

store_bikedata (city = 'nyc', bikedb = 'bikedb', dates = 201601:201603)

where the bikedb parameter provides the name for the database, and the optional argument dates can be used to specify a particular range of dates (Jan-March 2016 in this example). The store_bikedata function returns the total number of trips added to the specified database. The primary objects returned by the bikedata packages are ‘trip matrices’ which contain aggregate numbers of trips between each pair of stations. These are extracted from the database with:

tm <- bike_tripmat (bikedb = 'bikedb')
dim (tm); format (sum (tm), big.mark = ',')
#> [1] 518 518
#> [1] "2,019,513"

During the specified time period there were just over 2 million trips between 518 bicycle docking stations. Note that the associated databases can be very large, particularly in the absence of dates restrictions, and extracting these data can take quite some time.

Data can also be aggregated as daily time series with

bike_daily_trips (bikedb = 'bikedb')
#> # A tibble: 87 x 2
#>    date       numtrips
#>    <chr>         <dbl>
#>  1 2016-01-01    11172
#>  2 2016-01-02    14794
#>  3 2016-01-03    15775
#>  4 2016-01-04    19879
#>  5 2016-01-05    18326
#>  6 2016-01-06    24922
#>  7 2016-01-07    28215
#>  8 2016-01-08    29131
#>  9 2016-01-08    21140
#> 10 2016-01-10    14481
#> # ... with 77 more rows

A summary of all data contained in a given database can be produced as

bike_summary_stats (bikedb = 'bikedb')
#>    num_trips num_stations          first_trip       last_trip latest_files
#> ny  2019513          518 2016-01-01 00:00    2016-03-31 23:59        FALSE

The final field, latest_files, indicates whether the files in the database are up to date with the latest published files.

2.1 Filtering trips by dates, times, and weekdays

Trip matrices can be constructed for trips filtered by dates, days of the week, times of day, or any combination of these. The temporal extent of a bikedata database is given in the above bike_summary_stats() function, or can be directly viewed with

bike_datelimits (bikedb = 'bikedb')
#>              first               last 
#> "2016-01-01 00:00" "2016-03-31 23:59"

Additional temporal arguments which may be passed to the bike_tripmat function include start_date, end_date, start_time, end_time, and weekday. Dates and times may be specified in almost any format, but larger units must always precede smaller units (so years before months before days; hours before minutes before seconds). The following examples illustrate the variety of acceptable formats for these arguments.

tm <- bike_tripmat ('bikedb', start_date = "20160102")
tm <- bike_tripmat ('bikedb', start_date = 20160102, end_date = "16/02/28")
tm <- bike_tripmat ('bikedb', start_time = 0, end_time = 1) # 00:00 - 01:00
tm <- bike_tripmat ('bikedb', start_date = 20160101, end_date = "16,02,28",
                 start_time = 6, end_time = 24) # 06:00 - 23:59
tm <- bike_tripmat ('bikedb', weekday = 1) # 1 = Sunday
tm <- bike_tripmat ('bikedb', weekday = c('m', 'Th'))
tm <- bike_tripmat ('bikedb', weekday = 2:6,
                    start_time = "6:30", end_time = "10:15:25")

2.2 Filtering trips by demographic characteristics

Trip matrices can also be filtered by demographic characteristics through specifying the three additional arguments of member, gender, and birth_year. member = 0 is equivalent to member = FALSE, and 1 equivalent to TRUE. gender is specified numerically such that values of 2, 1, and 0 respectively translate to female, male, and unspecified. The following lines demonstrate this functionality

sum (bike_tripmat ('bikedb', member = 0))
sum (bike_tripmat ('bikedb', gender = 'female'))
sum (bike_tripmat ('bikedb', weekday = 'sat', birth_year = 1980:1990,
                   gender = 'unspecified'))

3. Citation

citation ("bikedata")
#> To cite bikedata in publications use:
#>   Mark Padgham, Richard Ellison (2017). bikedata Journal of Open Source Software, 2(20). URL
#> A BibTeX entry for LaTeX users is
#>   @Article{,
#>     title = {bikedata},
#>     author = {Mark Padgham and Richard Ellison},
#>     journal = {The Journal of Open Source Software},
#>     year = {2017},
#>     volume = {2},
#>     number = {20},
#>     month = {Dec},
#>     publisher = {The Open Journal},
#>     url = {},
#>     doi = {10.21105/joss.00471},
#>   }

4. Code of Conduct

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.




Minor changes:

  • add NEWS & README to CRAN description
  • Minor bug fixes


  • New helper function bike_cities to directly list cities included in current package version

Minor changes:

  • Bug fix for San Fran thanks to @tbdv (see issue #78)
  • Bug fix for LA (see issue #87)


  • Major expansion to include new cities of San Francisco, Minneapolis/St Paul, Montreal Canada, and Guadalajara Mexico
  • Most code restructured to greatly ease the process of adding new cities (see github wiki for how to do this).
  • New co-author: Tom Buckley

Minor changes:

  • Bug fix for LA data which previously caused error due to invisible mac OSX system files being bundled with the distributed data
  • More accurate date processing for quarterly LA data


  • Important bug in dodgr package rectified previously bug-ridden bike_distmat() calculations (thanks Joris Klingen).
  • Files for Washington DC Capital Bike Share system changed from quarterly to annual dumps
  • One rogue .xlsx file from London now processed and read properly (among all other well-behaved .csv files).
  • Update bundled sqlite3: 3.21 -> 3.22


  • New function bike_distmat() calculates distance matrices between all pairs of stations as routed through street networks for each city.
  • Helper function bike_match_matrices() matches distance and trip matrices by start and end stations, so they can be directly compared in standard statistical routines.
  • North American Bike Share Association (NABSA) systems (currently LA and Philly) now distinguish member versus non-member based on whether usage is 30-day pass or "Walk-up".

minor changes

  • dl_bikedata() function also aliased to download_bikedata(), so both do the same job.
  • Repeated runs of store_bikedata() on pre-existing databases sometimes re-added old data. This has now been fixed so only new data are added with each repeated call.
  • Dates for NABSA cities (LA and Philadelphia) are given in different formats, all of which are now appropriately handled.
  • Internally bundled sqlite3 upgraded v3.19 -> v3.21


  • Database no longer automatically indexed; rather indexes must be actively generated with index_bikedata_db(). This makes multiple usages of store_bikedata() faster and easier.
  • store_bikedata() fixed so it only unzips files not already in database (it used to unzip them all)
  • Internal changes to improve consistency (mostly through using the DBI package).


  • Minor changes only
  • More informative messages when data for specified dates not available


  • No change to package functionality
  • Drop dplyr dependency after dplyr 0.7 upgrade

0.0.1 (31 May 2017)

  • Initial CRAN release

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


0.2.4 by Mark Padgham, 6 months ago,

Report a bug at

Browse source code at

Authors: Mark Padgham [aut, cre] , Richard Ellison [aut] , Tom Buckley [aut] , Ryszard Szymański [ctb] , Bea Hernández [rev] (Bea reviewed the package for ropensci , see , Elaine McVey [rev] (Elaine reviewed the package for ropensci , see , SQLite Consortium [ctb] (Authors of included SQLite code)

Documentation:   PDF Manual  

GPL-3 license

Imports DBI, httr, lubridate, magrittr, methods, Rcpp, readxl, RSQLite, reshape2, tibble, xml2

Suggests dodgr, knitr, rmarkdown, roxygen2, testthat, covr

Linking to BH, Rcpp

System requirements: C++11

See at CRAN