Developing odbc

This vignette is intended to help developers of the R package install needed dependencies.

While the odbc package contains some documentation on how to install and configure database drivers in vignette("setup"), the documentation assumes that users are connecting to databases that have already been set up. In order to test package functionality, though, odbc sets up small example database deployments.

For the most part, this vignette assumes a MacOS system with aarch64 (e.g. M1 or M2) architecture. For Linux example code, see .github/workflows/db.yaml, and for Windows, see .github/workflows/db-windows.yml.

Posit Professional Drivers

Posit employees have access to the Posit Professional drivers and the infrastructure used to build and test them in the rstudio/pro-drivers GitHub repository. The Posit Professional drivers are a set of drivers vendored from Magnitude Simba that support many of the most popular DBMS, including SQL Server, Oracle, Redshift, Databricks, Snowflake, etc. The repository they’re developed in contains tooling to spin up a number of databases in docker containers to test against.

Note that Athena, Hive, Impala, MongoDB, and Oracle drivers are not available for macOS aarch64 (M1, M2, etc) at the time of writing.

Drivers

The only documented installation method for these drivers on MacOS is via RStudio Desktop Pro. The Posit confluence page titled “$$INTERNAL$$ Posit License Files for Employee Use” contains instructions for installing RStudio Desktop Pro. Once RStudio Pro is installed, install individual drivers as documented here.

Databases

Among other things, the rstudio/pro-drivers GitHub repository defines a MAKE tool for setting up and tearing down databases in docker containers. Ensure that you have a docker daemon running (i.e. Docker Desktop open) and, if you’re on macOS aarch64, have Settings > Use Rosetta for x86_64/amd64 emulation on Apple Silicon enabled. To start a container for a given dbms, run MAKE dist=none DB=dbms up, and tear it down with MAKE dist=none DB=db down. To see available dbms options, see the names of .yml files in the docker-compose directory. Find connection details for each database in docker/shared/odbc.ini.

DBMS-specific notes:

• sqlserver: Be sure to pass uid and pwd as arguments to dbConnect() explicitly, even though they’re in the odbc.ini as well.

On macOS aarch64, you will see a requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) warning when you start any of these containers. Feel free to ignore. :)

PostgreSQL

On MacOS, install PostgreSQL with:

brew install postgresql@14

You’ll also need to install and configure the PostgreSQL driver psqlodbc; see vignette("setup") to learn more.

To launch a PostgreSQL server locally, run:

brew services start postgresql@14

Next, create a database called “test” (or by whatever name is in the entry Database in your odbc.ini file):

createdb test

At this point, you should be able to connect to PostgreSQL through the R interface. Connect with:

postgres <- dbConnect(odbc(), "PostgreSQL")

where "PostgreSQL" is replaced with whatever DSN you’ve configured.

MySQL

First, installing MySQL with Homebrew:

brew install mysql@8.2

MariaDB drivers are compatible with MySQL and are more easily installed than MySQL drivers themselves in most cases. To install the MariaDB driver:

brew install mariadb-connector-odbc

Then, link the MariaDB driver with your MySQL data source name. That is, with the driver name [MariaDB] configuring your MariaDB install in odbcinst.ini, the first lines of your odbc.ini entry should look something like:

[MySQL]
Driver = MariaDB

After running brew services start mysql if needed, and confirming that the database is running with brew services info mysql, you should be able to:

library(odbc)
dbConnect(odbc(), "MySQL")

The second argument "MySQL" refers to the data source name configured above.

SQL Server test setup

To run Microsoft SQL Server on aarch64 (e.g. M1 or M2) MacOS, you will need:

• Docker 4.16 or higher
• MacOS 13 Ventura (or higher)

If needed, install Docker with:

brew install --cask docker

The Docker Desktop app provides a GUI to monitor deployed Docker containers and lives in Docker.app > Show Package Contents > Contents > MacOS > Docker Desktop.app.

To install the SQL Server ODBC driver and (optional) command line tool, use:

brew tap microsoft/mssql-release https://github.com/Microsoft/homebrew-mssql-release
brew install microsoft/mssql-release/msodbcsql18 microsoft/mssql-release/mssql-tools18

The odbc.ini entry should look something like:

[MicrosoftSQLServer]
driver = ODBC Driver 18 for SQL Server
Server = 127.0.0.1
port = 1433
Encrypt = no

With the driver name in odbcinst.ini being [ODBC Driver 18 for SQL Server].

With docker and the needed driver installed, deploy the container with:

sudo docker run \
  --platform linux/amd64 \
  -e "ACCEPT_EULA=Y" \
  -e 'MSSQL_SA_PASSWORD=BoopBop123!' \
  -p 1433:1433 \
   --name sql1 \
   --hostname sql1 \
   -d mcr.microsoft.com/mssql/server:2022-latest

The --platform tag is correct for M1; if you see Status: Exited (1) in Docker Desktop or a warning about incompatible architectures, navigate to Settings > General and ensure that Use Rosetta for x86/amd64 emulation on Apple Silicon is checked.

To connect via odbc, we need to pass the UID and PWD via the connection string; configuring those arguments via odbc.ini is not permitted. With the container deployed as above, the connection arguments would be:

con <- dbConnect(
  odbc::odbc(), 
  dsn = "MicrosoftSQLServer", 
  uid = "SA", 
  pwd = "BoopBop123!"
)

Then do some configuration of the server to add a testuser and create the test database

To configure a server to add a testing user and create a test database:

# Add a test user, but currently unused
dbExecute(con, "USE test")
dbExecute(con, "EXEC sp_configure 'contained database authentication', 1")
dbExecute(con, "RECONFIGURE")
dbExecute(con, "alter database test set containment = partial")
dbExecute(con, "CREATE USER testuser with password = 'BoopBop123!'")
dbExecute(con, "GRANT CONTROL TO testuser")
dbExecute(con, "DROP USER testuser")

# Create a test database
dbExecute(con, "CREATE DATABASE test")

On Linux, create a docker container with:

docker run -v "$(pwd)":"/opt/$(basename $(pwd))":delegated --security-opt=seccomp:unconfined --link sql1 -it rstudio/r-base:3.6.1-bionic /bin/bash

Then run:

curl https://packages.microsoft.com/keys/microsoft.asc | apt-key add -
#Ubuntu 18.04
curl https://packages.microsoft.com/config/ubuntu/18.04/prod.list > /etc/apt/sources.list.d/mssql-release.list

apt-get update
ACCEPT_EULA=Y apt-get install -y msodbcsql17
apt-get install -y unixodbc-dev

The resulting odbc.ini file will look something like:

[MicrosoftSQLServer]
driver = ODBC Driver 17 for SQL Server
Server = sql1
port = 1433
Database = test

SQLite

MacOS provides SQLite natively. With the SQLite odbc driver installed (see vignette("setup") if needed), run:

library(odbc)
dbConnect(odbc(), "SQLite")

The above example assumes the configured SQLite DSN is called "SQLite".

Oracle

A huge pain.

On MacOS

Getting an Oracle database booted up in a Docker container on MacOS is a bit challenging. First, you’ll build the container image yourself from the official Oracle source using the instructions here. Once you’ve finished building the image, run it with:

docker run -d --name oracle -p 1521:1521 -e ORACLE_PWD=odbcTest321 oracle/database:19.3.0-ee

Note that the run command here is a bit different than the one in the linked blog post.

Next, you’ll install the Basic, ODBC, and SQLPlus Instant Client libraries from the Oracle website (https://www.oracle.com/database/technologies/instant-client/macos-arm64-downloads.html) by doing the following:

• Download the Basic, ODBC, and SQLPlus .dmgs.
• Create a folder for the instant client to live in. I made mine at /Users/simoncouch/instantclient_23_3. (This folder will show up many times in the rest of these instructions–always substitute it with your own from here on out.)
• For each of the dmgs, open up the image and copy and paste its contents into the folder you made in the previous step.
• Symlink your libodbcinst.2.dylib (unixODBC library) into the folder. I used ln -s /opt/homebrew/Cellar/unixodbc/2.3.12/lib/libodbcinst.2.dylib /Users/simoncouch/instantclient_23_3/libodbcinst.2.dylib to do so for myself; use odbc_config --lib-prefix to help locate yours.

Make a text file at /Users/simoncouch/instantclient_23_3/network/admin/tnsnames.ora. Open that file (maybe with file.edit("/Users/simoncouch/instantclient_23_3/network/admin/tnsnames.ora")) and add the following:

db=
   (DESCRIPTION=
    (ADDRESS=(PROTOCOL=tcp)(HOST=0.0.0.0)(PORT=1521))
    (CONNECT_DATA=(SERVICE_NAME=orclpdb1)))

Now, set some paths to help R and the Instant Client communicate. In usethis::edit_r_profile(), add:

Sys.setenv(PATH = paste0("/Users/simoncouch/instantclient_23_3:", Sys.getenv("PATH")))
Sys.setenv(LD_LIBRARY_PATH = "/Users/simoncouch/instantclient_23_3")
Sys.setenv(ORACLE_HOME = "/Users/simoncouch/instantclient_23_3")
# this one is the default, but just to prevent gotchas:
Sys.setenv(TNS_ADMIN = "/Users/simoncouch/instantclient_23_3/network/admin")

(These are indeed all environmental variables that could be set in .Renviron. Be sure to substitute <$PATH> carefully if you choose to do so there.)

Note that there’s a MacOS-specific bug where RStudio overwrites PATH when set in Rprofile/Renviron before the 2024.04.0 “Chocolate Cosmos” release, so be sure to update your RStudio if needed.

In the terminal, test your Instant Client installation and database connection with:

sqlplus sys/odbcTest321@//localhost:1521/ORCLPDB1 as sysdba

If you’re able to connect, make a test user with:

CREATE USER test IDENTIFIED BY odbc;
GRANT DBA TO test;
exit;

Test connecting with that new user with:

sqlplus test/odbc@//localhost:1521/ORCLPDB1

If that succeeds, configure a DSN in your odbc.ini file with:

[Oracle]
Driver = /Users/simoncouch/instantclient_23_3/libsqora.dylib.23.1
Port = 1521
Host = 0.0.0.0

Then, connect in R with:

library(odbc)

con <- dbConnect(
  odbc(), 
  dsn = "Oracle", 
  DBQ = "db", 
  UID = "test", 
  PWD = "odbc"
)

Potential gotchas:

• Setting those DBQ, UID, and PWD parameters in my .ini file instead of the call to dbConnect() introduces errors for me.
• Note that, as of July 2024, there are no Posit Pro Drivers for Oracle on ARM Macs.
• If you see a “file not found” error for a driver that does exist, it’s because some dependency for the driver doesn’t exist; use otool -L /yada/yada/drivername.dylib to find the dependencies and then symlink them from elsewhere on your computer into where the driver looks for them.

On Linux

The process looks a bit different on Linux. These instructions are for a machine running Ubuntu 22.04.

Two steps here–1) booting up an Oracle database in a docker container and 2) installing and configuring the Oracle Instant Client and driver to connect to that database.

If you haven’t already, you’ll need to install Docker Desktop, unixODBC, and libaio1. You can get the last with:

sudo apt-get install libaio1

First, to boot up an Oracle database in a Docker container, run:

docker run -d \
  --name oracle2 \
  -p 1521:1521 \
  -e ORACLE_RANDOM_PASSWORD=true \
  -e ORACLE_DATABASE=test \
  -e APP_USER=RodbcR \
  -e APP_USER_PASSWORD=Password12 \
  gvenzl/oracle-xe:21.3.0

Some documentation on this Docker image here.

Now, as for installing and configuring the Oracle Instant Client and driver, the PRs implementing support for unit testing Oracle and adding support for the Oracle Pro Driver for odbc are a great guide here. The “Installation Instructions” at the bottom of Oracle’s Instant Client docs (https://www.oracle.com/database/technologies/instant-client/linux-x86-64-downloads.html) are thorough–you shouldn’t need them here, but a good backup.

I ultimately wasn’t able to make a connection with the driver that Oracle ships, but was able to do so with the Posit Pro Drivers.

Run the lines in .github/odbc/install-oracle-driver.sh from the first linked PR, which installs both the Instant Client and Oracle’s driver.

Add the following lines to your .Rprofile (with usethis::edit_r_profile()):

Sys.setenv(LD_LIBRARY_PATH = paste0("/opt/oracle/instantclient_21_12:", Sys.getenv("LD_LIBRARY_PATH")))
Sys.setenv(PATH = paste0("/opt/oracle/instantclient_21_12:", Sys.getenv("PATH")))

Create a text file called /opt/oracle/instant_client_21_12/tnsnames.ora and paste the following inside of it:

TEST=
   (DESCRIPTION=
    (ADDRESS=(PROTOCOL=tcp)(HOST=0.0.0.0)(PORT=1521))
    (CONNECT_DATA=(SERVICE_NAME=test)))

Install the RStudio Pro Oracle driver, either using the installation code from the linked PR (see the repository secret for the driver URL or contact Simon Couch or Hadley Wickham if you don’t have access) or through RStudio Pro (if you have credentials, you can generate yourself a license key here: https://connect.posit.it/support/license_management/).

Add the following lines to your .Renviron (with usethis::edit_r_environ()). Adjust the value for ODBCSYSINI to the directory prefixing .ini files in odbcinst -j.

TNS_ADMIN=/opt/oracle/instantclient_21_12
ODBCSYSINI=/etc
ORACLE_HOME=/opt/oracle

Symlink your Instant Client install into the Pro Driver install directory with sudo ln -s /opt/oracle/instantclient_21_12/* /etc/pro_drivers/oracle/bin/lib/.

Add the following to your odbc.ini:

[ProOracle]
Driver = /etc/pro_drivers/oracle/bin/lib/liboracleodbc_sb64.so
Port = 1521
Host = localhost
SVC = FREEPDB1
UID = RodbcR
PWD = Password12
DBQ = test

Now, putting the pieces together, connect to the Oracle Database docker container from R with:

library(odbc)
dbConnect(odbc(), dsn = "ProOracle")

Snowflake

To obtain a Snowflake account:

• If you’re a Posit employee, reach out to James Blair.
• If you’re not affiliated with Posit, sign up for a free trial.

Install drivers using the instructions on the Snowflake documentation. For MacOS, feel free to ignore sections noting that iODBC is required.

On MacOS and Linux:

• Locate your simba.snowflake.ini file, likely in opt/snowflake/. You will need sudo access to edit it.
  • Change DriverManagerEncoding to UTF-16.
  • Change the uncommented ODBCInstLib entry to the path of your unixODBC install.
• In the Snowflake UI, navigate to Worksheets, make a new worksheet using the blue “+” button in the top right-hand corner, and run the line SELECT SYSTEM$ALLOWLIST(). In the output, copy the URL corresponding to "type":"SNOWFLAKE_DEPLOYMENT". This Deployment URL should look like a random string, an AWS region, and the snowflake API URL, e.g. wmc39401.us-east-1.snowflakecomputing.com.

The odbc.ini file should look like:

[Snowflake]
Driver      = Snowflake
Description =
uid         = <uid>
server      = <deployment-url>

The above assumes Driver = Snowflake is configured in odbcinst.ini. Then, connect by supplying the password as an argument:

con <- dbConnect(odbc(), dsn = "Snowflake", pwd = Sys.getenv("snowflakePass"))

In the above, the password has been set as Sys.setenv(snowflakePass = "actualPass").

Amazon Redshift

If you’re a Posit employee, you should have access to the rstudio/warehouse GitHub repository. Follow the instructions there to get access to the internal Redshift cluster. Access to the cluster is only enabled through Posit’s internal Workbench instance, where the professional drivers will already be installed.

Note that Redshift is based on a modified version of PostgreSQL.

RODBC

We need to install the RODBC package for benchmarking in the vignette vignette("benchmarks"). The CRAN version of RODBC uses iODBC, so to use unixODBC we need to recompile it from source, specifying the odbc manager explicitly:

install.packages("RODBC", type = "source", INSTALL_opts="--configure-args='--with-odbc-manager=odbc'")