Images
Creating a reproducible environment for data science is a common challenge when working as a team. It is common practice to put several install commands at the beginning of a notebook or script to help ensure that others will be able to recreate your analysis. Unfortunately, this installation step can take a long time when many different libraries and packages are required.
This is where Zepl Images can help you and your team save significant time. Every time you run a notebook, you are spinning up a container which contains an "image", which is a prebuilt environment with all the libraries and settings your team needs for reproducible data science. Images lets you create Images that contain as many or as few libraries as your team needs.
Contact [email protected] to get access and learn more!
Images can easily be found under Resources. To start creating a new image, simply click "CREATE NEW". If you have another Image which you’d like to modify as a template for your new Image, you can also "Clone" that template under the actions for that Image.
Each Image should have a unique, descriptive name that makes it easy to find later. As a first step of creating an Image, give your image a name and a description of what you plan to use it for.
After naming your image, you are ready to start configuring your Image with interpreters, libraries, packages, and system dependencies. The syntax for doing this is documented below.
To build your image, click the "Create" button at the bottom. Building the image will take a few minutes, and takes longer when more libraries are installed to the image. Sometimes builds can fail if the wrong syntax is used, or there is a library version mismatch issue. If that occurs, you can download a log in the image actions which describes where the build failed. If the image builds successfully, congrats! You can now attach it to notebooks. If you’d like the image you created to be the standard for your entire organization, you can set it as the default image in the image actions. You can also edit or delete an image under these image actions.
To add a Python interpreter to an Image, simply click "+ Python" in the image creation screen. Zepl supports Python 3.8 as the default Python interpreter and can be referenced by the %python alias,
To add libraries to your Image, you can use the following supported to install libraries using pip:
Librarynamelibraryname==version
In addition to any libraries that you include in your Image list, Zepl installs multiple Python libraries to the image necessary for Zepl to function well. If you try to install a different version of these libraries, it is possible you may run into a compatibility issue. You can always see what versions of libraries you have installed in your image by running a pip list command in the notebook. The following is the syntax for doing so:
%python!pip list
As of July 2020, the list of libraries we install for Python in an image are the following:
grpcio==1.24.3ipykernel==5.1.3ipython==7.13.0jupyter-client==5.3.4protobuf==3.10.0py4j==0.10.8.1boto3==1.10.9cassandra-driver==3.22.0google-cloud-bigquery==1.21.0mysql-connector-python==8.0.18snowflake-connector-python==2.2.7hdbcli==2.4.167psycopg2-binary==2.8.4pyodps==0.8.4
To add an R interpreter to an Image, simply click "+ R" in the interpreter creation screen. Zepl supports R 3.6 as the default R interpreter for an Image and can be referenced with the %r alias.
To add libraries from a CRAN server to your Image, you simply list out the libraries you hope to install on separate lines. This installs packages similarly to using install.package("LibraryName") in the notebook
Libraryname1Libraryname2
In addition, you can install libraries using the devtools. The following syntaxes are supported:
devtools::install_github() from githubdevtools::install_bitbucket() from bitbucketdevtools::install_url() from an arbitrary urldevtools::install_version() installs a specified version from cran
In addition to any libraries that you include in your Image list, R comes with many libraries already installed, and Zepl installs multiple R libraries to the image necessary for Zepl to function well. If you try to install a different version of these libraries, it is possible you may run into a compatibility issue. You can always see what versions of libraries you have installed in your image by running an installed.packages() command. As of July 2020, the list of libraries we install for R in an Image are the following:
devtoolsDatabaseConnectordplyrknitrrJavaRJDBCsnowflakedb/dplyr-snowflakedb
To add a Spark interpreter to an Image, simply click "+ Spark" in the interpreter creation screen. Zepl supports Spark 2.3.2 as the default Spark interpreter for Images, which can be referenced with the %spark alias.
To add libraries to Spark, you can do so by adding maven dependencies on separate lines. Please note that we only support compile and exclude_group commands in the syntax today.
compile "commons-io:commons-io:2.6"compile("org.apache.hadoop:hadoop-aws:2.8.3") { exclude group ‘javax.servlet’, module: ‘servlet-api’}
In addition to any libraries that you include in your Image list, Spark comes with many libraries already installed, and Zepl installs multiple Spark libraries to the Image necessary for Zepl to function well. If you try to install a different version of these libraries, it is possible you may run into a compatibility issue. You can always see what versions of libraries you have installed in your Image by running the following code in the notebook:
%pyspark!ls /usr/zepl/interpreter/lib/
As of July 2020, the list of libraries we install for Spark in an Image are the following:
interpreterDatasource "mysql:mysql-connector-java:8.0.18"interpreterDatasource "org.postgresql:postgresql:42.2.8"interpreterDatasource "net.snowflake:snowflake-jdbc:3.12.5"interpreterDatasource "com.sap.cloud.db.jdbc:ngdbc:2.4.63"interpreterDatasource "com.aliyun.odps:odps-jdbc:3.1.0"interpreterDatasource "com.google.cloud.spark:spark-bigquery-with-dependencies_2.11:0.15.1-beta"interpreterDatasource "com.aliyun.odps:odps-jdbc:3.1.0"interpreterDatasource("com.datastax.spark:spark-cassandra-connector_2.11:2.0.11")interpreterDatasource("net.snowflake:spark-snowflake_2.11:2.7.1-spark_2.2")interpreterDatasource("org.apache.spark:spark-hive_2.11:2.2.1")
You can add environmental variables and packages to your Image with this feature. This can be helpful because certain libraries sometimes require underlying infrastructure to be installed to or environment variables set in the Image for certain libraries to work properly.
You can set an environment variable with very simple syntax:
VAR1=123VAR2=456
You can install packages from GitHub directories with Maven syntax.
Images can easily be managed and analyzed from the Images management console. For any existing Image that you have created, you can see what it is called, who created the Image, if the Image successfully created, when it was created, and what notebooks are associated with the Image.
For any Image that you are curious in learning more about, you can click on the row to learn more about how that Image was constructed. Under the actions menu for Images, you can perform the following operations:
Clone: This lets you use an existing Image as a template to modify while creating a new Image
Edit: This lets the creator of an Image modify the configuration of the Image
Delete: This lets the creator of an Image delete the Image
Set as Default: This sets an Image as the default Image for all users in an organization
When you are creating a new notebook, you must select the Image you want to use to execute code from that notebook.
In the notebook, we’ve designed a toolbar on the right hand side which helps you see what interpreters and libraries are available for you to use in your Image.
You can change this selection at any time for a given notebook by editing the notebook settings. To do so, your container must be shut down.
