This is an SDK for using the Looker API with the R programming language. Our priority is to add support for the endpoints that are most useful to data scientists and analysts working in R (the ones that get data from Looker) but eventually the goal is to add support for every endpoint in the Looker API.
For more information about the Looker API’s capabilities and for details about what each endpoint does, see the Looker API reference documentation.
This SDK is NOT officially supported by Looker/Google. It is community supported. Issues can be logged here in the GitHub Issues page. Details of Looker API and SDK support can be found at https://cloud.google.com/looker/docs/api-sdk-support-policy.
LookR used to be based on
rJava
, but this caused
problems with reliability and many users struggled to get it to work. To
fix this, we’ve rewritten a new SDK in native R. The versions have been
reset as well.
All of the old code is available on the old-lookr branch, so if you want
to continue using it, you an install from that branch by setting* ref
to old-lookr
*:
devtools::install_github("looker/lookr", ref = "old-lookr")
The code recently changed to support API 4.0 instead of 3.0. There
still is not a robust testing mechanism in place. If you need to use the
older version install it from the api-3.0
branch.
devtools::install_github("looker/lookr", ref = "api-3.0")
# or
devtools::install_github("looker/lookr", ref = "api-3.1")
Use devtools
to install from this repository:
devtools::install_github("looker/lookr", ref = "api-4.0")
library(lookr)
First, make sure that you have your API credentials and the URL to your
Looker host in a file called looker.ini
. LookR uses the information in
this file to authenticate. See below for a sample of what looker.ini
should look like:
looker-sample.ini
[Looker]
# API version is required
api_version=4.0
# Base URL for API. Do not include /api/* in the url
base_url=https://<your-looker-endpoint>:19999
# API 3 client id
client_id=your_API3_client_id
# API 3 client secret
client_secret=your_API3_client_secret
# Optional embed secret for SSO embedding
embed_secret=your_embed_SSO_secret
# Optional user_id to impersonate
user_id=
# Set to false if testing locally against self-signed certs. Otherwise leave True
verify_ssl=True
Make sure that this file is not checked into version control.
All of the endpoints are accessed from methods contained in an instance
of the LookerSDK
class, which is implemented in
R6
. To start
using LookR, create an instance of that class:
sdk <- LookerSDK$new(configFile = "looker.ini")
If you’re using a version of RStudio that includes the Connections Pane, you can also use that to create a new connection to Looker.
Whether you use the Connection Pane to create the connection or not, you will be able to browse your projects, models, and explores in the Connections Pane as long as your API user has access to them.
You can use LookR to fetch the results of a saved Look by calling
runLook
:
data <- sdk$runLook(lookId = 1337)
You can also use a new query by calling runInlineQuery
. model
,
view
, and fields
are mandatory arguments:
data <- sdk$runInlineQuery(model = "model_name",
view = "my_awesome_explore",
fields = c("id", "count"))
and you can optionally supply filters
, sorts
, limit
, and
queryTimezone
:
data <- sdk$runInlineQuery(model = "model_name",
view = "my_awesome_explore",
fields = c("id", "count"),
filters = list(field.name = jsonutil::unbox("filter_expression"),
another.field = jsonutil::unbox("filter_expression")),
limit = 500,
queryTimezone = "America/Los_Angeles")
The SDK returns data as a list of lists. The inner list has named elements for one row of data. The outer list is the collection of rows. R most naturally wants to work with columns of data. The following process will reformat the data in a useful way…
sdk <- lookr::LookerSDK$new(configFile = "looker.ini")
# run a look or inline query
data <- sdk$runLook(lookId = 5 )
# Verify that at least one row was returned
length(data)
[1] 5
# get the column names from the first row
names(data[[1]])
[1] "customer.first_name" "order.count"
# create a new list that represents each column
fname = sapply(data, "[[", 1)
ocount = sapply(data, "[[", 2)
# create a data.frame object with the columns
# the list names will become the default column names
df <- data.frame(fname, ocount)
print(df)
fname ocount
1 Homer 5
2 Bart 4
3 Robert 3
4 Bernard 2
5 Milhouse 1
# If a column can have a NULL value, use I(...) to wrap the column list
ocount = list(5,4,NULL,2,1)
df <- data.frame(fname, I(ocount))
print(df)
fname ocount
1 Homer 5
2 Bart 4
3 Robert
4 Bernard 2
5 Milhouse 1
The SDK exposes a number of the most frequently used and useful operations But the whole API is available through underlying methods. For example:
sdk$refresh()
data <- sdk$userSession$lookApi$look(6, config = sdk$oauthHeader)$content
print(data)
<LookWithQuery>
Public:
...
folder_id: 5
fromJSON: function (LookWithQueryJson)
fromJSONObject: function (LookWithQueryJsonObject)
fromJSONString: function (LookWithQueryJson)
...
id: 6
...
model: LookModel, R6
public: NULL
public_slug: 2ytjPWrQrbxSHHZxGSYXXhgsF7Pjj6Wy
...
query: Query, R6
query_id: 124
short_url: /looks/6
space: SpaceBase, R6
space_id: 5
title: Simple Look
...
user_id: 1
view_count: 1
The sdk$refresh()
call checks if the session is still authenticated
and logs in again if needed.
The sdk
object has a userSession
attribute. The userSession
has
references to subsections of the API. For example, the lookApi
has the
operations that deal with Looks. The look
operation gets a Look by id.
config = sdk$oauthHeader
has to be the last parameter sent to any
of these calls. That contains the token that needs to be sent to
validate your authorization.
These calls return a responseObject
with two attributes.
responseObject$response
is the http response codes. Check the value of
httr::status_code(responseObject$response)
for 200
to verify
success. The responseObject$content
is the data that is actually
returned, assuming everything worked properly, or an error message.
Currently the UserSession
supports several but not all the API
objects.
R has tab completion to help explore these:
sdk$userSession$<TAB>
sdk$userSession$.__enclos_env__ sdk$userSession$lookmlModelApi sdk$userSession$clientConfig sdk$userSession$activeToken
sdk$userSession$oauthHeader sdk$userSession$userApi sdk$userSession$settings sdk$userSession$me
sdk$userSession$userToken sdk$userSession$dashboardApi sdk$userSession$clone sdk$userSession$isAuthenticated
sdk$userSession$token sdk$userSession$queryApi sdk$userSession$logout sdk$userSession$loginUser
sdk$userSession$tokenExpiresAt sdk$userSession$lookApi sdk$userSession$impersonating sdk$userSession$login
sdk$userSession$userId sdk$userSession$authClient sdk$userSession$setAuthExpiration sdk$userSession$reset
sdk$userSession$projectApi sdk$userSession$apiClient sdk$userSession$assignApiClient sdk$userSession$initialize
We can see that UserSession
has references to the projectApi
,
userApi
, queryApi
, dashboardApi
, and lookApi
.
If we want to access an Api that isn’t readily available, we can do it like this:
workspaceApi <- lookr::WorkspaceApi$new(sdk$userSession$apiClient)
workspaceApi$ <TAB>
workspaceApi$.__enclos_env__ workspaceApi$userAgent workspaceApi$workspace workspaceApi$initialize
workspaceApi$apiClient workspaceApi$clone workspaceApi$all_workspaces
data = workspaceApi$all_workspaces(config = sdk$oauthHeader)$content
print(data)
[[1]]
<Workspace>
Public:
can: list
clone: function (deep = FALSE)
fromJSON: function (WorkspaceJson)
fromJSONObject: function (WorkspaceJsonObject)
fromJSONString: function (WorkspaceJson)
id: production
initialize: function (can, id, projects)
projects: list
toJSON: function ()
toJSONString: function ()
[[2]]
<Workspace>
Public:
can: list
clone: function (deep = FALSE)
fromJSON: function (WorkspaceJson)
fromJSONObject: function (WorkspaceJsonObject)
fromJSONString: function (WorkspaceJson)
id: dev
initialize: function (can, id, projects)
projects: list
toJSON: function ()
toJSONString: function ()