Week Ending 12/22/2017
- Added highlighting to the Workflows execution history panel to indicate which execution is currently being viewed.
- When creating a new task from the graph view in Workflows, the task editing panel will automatically open.
- Fixed a bug that was preventing published custom Scripts from being unpublishable.
- Fixed a bug in scheduled Workflows that was causing errors with overlapping execution schedules.
Week Ending 12/15/2017
- Past Query results are now collapsed by default to improve page performance.
- Changed how API requests are handled so that background requests will be dropped when the system is overloaded to accelerate recovery.
- Updated the Docker image for R scripts to datascience-r 2.3.0. New R scripts created in Platform now have the Civis R API Client installed automatically.
- Fixed a bug in Query autocomplete that was causing text to be inserted incorrectly into the query.
Week Ending 12/8/2017
- Updated the look of the user details page to reduce white space.
- Fixed a bug in Python script cloning so that the script’s required resources are also cloned.
- Fixed a bug in File imports that was causing errors when overriding column names that were initially duplicates.
- Fixed a security vulnerability in Notebooks.
Week Ending 12/1/2017
- Upgraded datascience-python Docker image to 3.3.0. New Python Scripts and Python Notebooks will use this image by default.
- New packages:
- civisml-extensions 0.1.5
- dask 0.15.4
- s3fs 0.1.2
- Package Updates
- civis 1.6.2 -> 1.7.1
- boto3 1.4.4 -> 1.4.5
- matplotlib 2.0.2 -> 2.1.0
- numpy 1.13.1 -> 1.13.3
- pandas 0.20.3 -> 0.21.0
- pyarrow 0.5.0 -> 0.7.1
- scikit-learn 0.19.0 -> 0.19.1
- cloudpickle 0.3.1 -> 0.5.1
- muffnn 1.1.2 -> 1.2.0
- pubnub 4.0.12 -> 4.0.13
- tensorflow 1.2.1 -> 1.4.0
- New packages:
- Moved conda to version 4.3.30.
- Improved the performance of job indices by loading user lists when opening the sharing pane rather than when the index page loaded.
- Added ability to set Memory and CPU allocations for Python and R Scripts.
- Fixed a bug that was allowing Queries to be executed in browser after a session timeout.
- Fixed a bug in Query autocomplete to allow autocompletion on table names that include numbers.
- Fixed a bug when validating schema and table names in Imports.
- Fixed a bug in Imports so that a new table is not created if there is an error in the Import.
Week Ending 11/17/2017
- Improved the performance of queries executed in the UI by moving API-initiated queries to separate resources.
- Fixed a bug in Workflows where an execution would fail if a job's error message container non-ASCII characters.
Week Ending 11/10/2017
- Python scripts now use v3-series datascience-python images. v3 includes the Civis Python API client v1.5.2 which includes CivisML capabilities.
- Added the ability for users to change when their Notebook will time out.
- Added the ability to set CPU, memory, and disk space on Python and R scripts.
- Changed the vertical height of Tableau reports accessed using public links so it matches how the report looks in Platform.
- Changed export encoding behavior so it doesn’t cause Postgres export jobs to fail.
- Increased the memory available for Database Export jobs.
- Improved the performance on all index pages.
- Fixed a bug where a workflow wasn’t marking tasks as complete after they finished.
- Fixed an error where queries were sometimes exposing the ruby error instead of the SQL error.
- Fixed issues with exporting data to Vertica.
- Fixed a bug where user admins could not see inactive users.
- Fixed incorrect warnings that a user’s platform session had timed out.
- Fixed a bug where the timezone wasn’t showing up in the automation pane.
- Fixed a bug where fullscreen reports loaded with blank white space on the right side of the screen.
Week Ending 11/3/2017
- Changed Python script editor to convert all tabs to spaces so as to better align with recommended style conventions.
- Performance improvements for index page filtering.
- Fixed bugs causing errors when cloning certain types of scripts.
Week Ending 10/27/2017
- Added support for using Redshift's TIMESTAMPTZ (time stamp with time zone) data type.
- Improved the formatting of EXPLAIN plans in Query to make it more readable.
- Changed the error message for Container Scripts to indicate when a run failed due to memory constraints.
- Changed the icon on Workflow tasks with errors to be an exclamation point instead of an "X" to be consistent with other Civis icons.
- Fixed a bug where SQL Scripts that resulted in a null result set were linking files from the last run that contained results.
- Fixed a bug where a user running a job owned by a different user could cause a job queue.
- Fixed a bug in Workflows that was causing a locked Workflow to appear editable after an execution.
Week Ending 10/20/2017
- Added the ability to use numbers in Civis usernames.
- Added the ability for run outputs to be set during script shutdown so the script writer can store debug information about the job.
- Added helpful descriptive text to the advanced options section of File Imports.
- Changed the width of tabbing in the Python editor from 2 to 4 characters to align with recommended style conventions.
- Added an error message to Container Scripts to indicate when the command input exceeds the character limit.
- Added a default value of 0 to the max errors field in File Imports.
- Fixed a bug in Database Imports that was preventing connections to external clusters.
Week Ending 10/13/2017
- New “Import from S/FTP” and “Import from URL” jobs can now be created from the “New Import” menu and previously created jobs can be found in Imports.
- New “Export to S/FTP” jobs can now be created from the “New Export” menu and previously created jobs can be found in Exports.
- SQL Scripts are no longer listed as Exports. They only appear on the Scripts index now.
- Fixed a bug in two-factor authentication that allowed 2FA code reuse under certain conditions.
- Fixed bugs in Database Exports to Vertica where incorrect syntax was being used to create the destination table.
- Fixed a bug in Notebooks where custom packages were not preserved when a notebook was cloned.
Week Ending 10/6/2017
- Added a new script parameter type of Multi line String to make it easier to format long blocks of text.
- Improved the performance of an import's page when listing source and destination databases.
- Fixed a bug in Salesforce and Database Imports where errors were not being properly surfaced.
- Fixed a bug in File Imports where column name overrides weren't working on files without header rows.
- Fixed a bug where Report Only users were able to see the Enhancements index.
- Fixed a bug in Database Imports where the source or destination doesn't have a schema.
Week Ending 9/29/2017
- Upgraded Container Scripts to use the datascience-python Docker image version 3.2.0.
- scikit-learn 0.18.2 -> 0.19.0
- civis 1.6.0 -> 1.6.2
- requests 2.14.2 -> 2.18.4
- urllib3 1.19 -> 1.22
- Improved the performance of filtering on the homepage activity filter.
- Improved the performance of the Models index.
- The data pane will quickly reflect changes made by additional ALTER SCHEMA commands.
- Added detailed logging to Imports and Export to better communicate the progress of the job.
- Added a clearer indication to File Imports that column names are editable.
- Fixed a bug in Address Correction where the output table name didn't appear if the output schema had no tables in it.
- Fixed a bug when searching for tables in Database Imports.
- Fixed a bug that was preventing Exports from being removed from a Project.
- Fixed a bug where Container Scripts built from a template were inaccurately routing the URL after closing the script.
- Fixed a bug that allowed non-administrators the ability to view the details of inactive users in their organization.
- Fixed a bug in Database Imports to enable the import of ENUM data types and columns with the UNSIGNED designation.
Week Ending 9/22/2017
- Schemas without any tables in them now show up in the data pane, are re-scanned when changed using queries, and can be selected in schema dropdowns.
- The “Export Results” link in Query now links to scripts page instead of jobs page to make it easier to download query results.
- Improved the performance of run history pane by eliminating unnecessary API calls.
- Code windows automatically expand as text lines increase to avoid having a scroll bar within the code window.
- Fixed a bug in scripts that allowed users to accidentally and irreparably delete text from the code window.
- Fixed a bug that was creating an unnecessary error message in sync jobs.
- Fixed a bug in scheduling custom scripts.
Week Ending 9/15/2017
-
Upgraded Civis Python API Client to v1.6.2
- Added ResourceWarning for Python 2.7.
- Added TypeError for multi-indexed dataframes when used as input to CivisML.
- ModelPipeline.from_existing will warn if users attempt to recreate a model trained with a newer version of CivisML, and fall back on the most recent prediction template it knows of.
- Make the PaginatedResponse returned by LIST endpoints a full iterator, which also makes the iterator=True parameter work in Python 2.
- When using civis.io.civis_to_csv, emit a warning on SQL queries which return no results instead of allowing a cryptic IndexError to surface.
- Fixed the example code snippet for civis.io.civis_to_multifile_csv and added more details on its return dict in the docstring.
- Pinned down sphinx_rtd_theme and numpydoc in dev-requirements.txt for building the documentation.
- Updated error message when creating batch imports to be more informative.
- Improved the performance of the CASS autosave feature.
- Change group selection in user pages to typeahead search.
- Removed the “active” checkbox from user creation page.
- Disabled the editability of email field in user detail page to match API behavior.
- Fixed a bug in CASS/NCOA jobs that was overwriting existing triggers.
Week Ending 9/8/2017
- Replaced the error displayed with a warning when a download callback query made via the Civis API client returns an empty result.
- When users call a list endpoint with the iterator argument from the Civis API client the response is now an iterable object.
Week Ending 9/1/2017
- Added an expires_at parameter to GET /files/:id to enable generated file links to last up to 36 hours.
- Changed the error message when importing ZIP files to be more informative.
- Fixed a bug in imports that was causing duplicate column errors.
Week Ending 8/25/2017
- Upgraded CivisML to v1.1.1
- Fixed a bug which prevented users from grid searching over a calibrated classifier model.
- Improve handling of tables containing a column with "date" type by now correctly reading SQL "date" type as a string.
- Added the ability for tables to have integer column names.
- Improved the handling of errors raised in prediction child jobs.
- Enabled retries for failed prediction child jobs.
- Improved the performance of the homepage by restricting it to only show recent activity (from the last 30 days).
- When creating a new table via imports, the primary key column is now wrapped in double quotes to support column names with spaces.
- Fixed a bug where scripts that contained a GRANT...ON ALL TABLES IN SCHEMA... SQL statement were failing.
- Fixed a bug that prevented users from being able to clone imports.
- Fixed a visual bug in the primary key field in the Score Model job where the "X" icon overlapped with the key name.
Week Ending 8/18/2017
- Updating Google authentication on an export will now redirect back to the export rather than an import.
- When importing a file with over 1,600 columns (Redshift's max column count), the error will now be properly displayed.
- Added a “CreateStatement” tab to the Table Details page that shows the SQL table definition (DDL) making it easier to create duplicate or modified versions of the table.
- Fixed a typo in the New Script pane.
Week Ending 8/11/2017
- Added tooltips to the left side navigation icons when the navigation pane is collapsed.
- Improved job failure messages by replacing System Errors with the full error.
- Implemented minor UI changes to the New Import and New Script side panels.
- Fixed a bug in the modeling wizard when selecting the sparse continuous option.
Week Ending 8/4/2017
-
Upgraded CivisML to v1.1
- Added the ability to install custom dependencies when using CivisML via the API client.
- Added null indicator column to categorical expansion to more correctly handle missing values in features.
- Improved cross-validation to perform better with ordered datasets.
- Improved the reliability of model training when encountering AWS errors.
- Improved the reliability of model scoring when fetching results from successful runs.
- Fixed a bug where modeling would fail if column names were reserved Redshift words.
- Links from the Reports index page will now open the report in full-screen mode
- Updated Imports text to reflect available data sources and direct users to help documentation
Week Ending 7/28/2017
- Added a "note" parameter to the /scripts/templates API endpoint. Users can now use POST or PATCH requests to add custom descriptive text to the top of a script template.
- Increased the maximum number of displayed run outputs per job run to 10. This will accommodate additional outputs of CivisML jobs.
- Closing a script built from a template will now properly take you to the Scripts index.
- Fixed a rare bug in Export to Google Sheet where the entire result set wasn't getting exported.
Week Ending 7/21/2017
- Improved the front-end performance of job status updates.
- Improved the message for job credential permission errors to be more descriptive.
- Improved the handling of canceled jobs in the Python API Client.
- Added support for tables and schemas showing up quickly in the data pane after adding a Redshift user to a Redshift group or running a GRANT ... ON ALL TABLES IN SCHEMA ... command.
- Fixed a bug in how credentials are used when running a script built from a template.
- Fixed a bug in the retry logic of file uploads.
- Fixed a bug in saving user preferences on index pages.
Week Ending 7/14/2017
- Improved the person matching algorithm when including email as a match field.
- Changed the default If Table Exists setting to "Fail" on Google Sheet imports to maintain consistency with other imports and to default to the most conservative option.
- Reworded job status notification emails to lead with the status to better communicate the results of jobs with long names.
- Fixed a bug where CivisML models were linking to the incorrect script page.
Week Ending 7/7/2017
- Files added to a Project will now appear in the UI. Files can be downloaded or removed from the project directly from the project page.
- Improved the performance of table caching in CivisML when starting multiple model scoring jobs from the same table.
- Fixed a bug in API documentation generation.
Week Ending 6/30/2017
- Updated GET /scripts documentation to include "containers" as a type parameter value.
- Fixed a bug in imports to remove stray period in the destination schema.table field.
Week Ending 6/23/2017
- Improved the performance of the Models index.
- Added clarification on the File Import page about which file types are supported.
- Fixed a bug where the first run of a Database Import could fail.
Week Ending 6/16/2017
- Improved the error message on File Imports when not specifying a destination schema.
- Fixed a bug in Export to Google Sheet jobs where the SQL code could prevent users from editing the job.
Week Ending 6/9/2017
- Added a helpful error message when a user attempts to add an invalid phone number to a Civis account.
- Improved the Civis - Tableau integration for more reliable automatic importing of Tableau reports to Civis.
- Fixed a bug in email triggers.
- Fixed a bug in displaying SQL views with TEXT type columns.
- Fixed a bug in the Google Sheets integration to reduce error messages when attempting to access a Google Sheet owned by someone else.
Week Ending 6/2/2017
- Fixed a bug when applying secondary groups in user creation.
- Fixed a bug to ensure Redshift credentials are added when adding a new user.
- Fixed a bug in credential switching when cloning an Export to Google Sheet.
- Fixed a bug in job cancelation to speed up cancel response time.
- Fixed a bug in canceling Import syncs.
Week Ending 5/26/2017
- Updated matching algorithm to improve table-to-table match rate involving zip codes.
- Improved the performance of the data pane.
- Added additional logging to Report Snapshot jobs to improve troubleshooting.
- Performance improvements on the Query page.
- Fixed a bug to allow viewing of run logs.
- Fixed a bug so only SQL scripts can be linked to reports.
- Fixed a bug in delimiter toggle for file imports.
- Fixed a bug in table scanner column count.
Week Ending 5/19/2017
- Enabled users to add a credential as a run output for a job via the API.
- Migrated all Google Sheet Import syncs into the updated Imports UI.
- Added a cancel_timeout parameter to Python, R, and Container scripts to allow users to insert buffer time for a container’s jobs to be canceled before the container is stopped
- Updated API documentation for Export to Google Sheets and Run Outputs to add clarity.
- Updated run log transfer to replace all characters outside of the Latin-1 character set with question marks to avoid an error while sending the logs via API.
- Enabled the data pane to be widened the full width of the window to accommodate long schema and table names.
- Performance improvements when listing custom scripts derived from a specific templateId.
- Performance improvements on the Exports and Reports index pages.
- Performance improvements on the home page.
- Fixed a bug in Google API rate limits.
- Fixed a bug in table details Javascript.
- Fixed a bug that was incorrectly labeling run output files as expired.
Week Ending 5/12/2017
- Released CivisML replacing our old modeling architecture and released support for Gradient Boosting Trees, Random Forests, and Extra Trees algorithms. Read about using CivisML in a script template here.
- Added clarity to acceptable file types for file imports.
- Added better visual confirmation during table scan after initial file import.
- Added pagination to the /jobs endpoint to prevent a request from overloading web servers.
- Added expanded error logs to notification emails.
- Fixed a bug in the Assume Role feature for admin accounts.
- Fixed a bug in table UNLOAD SQL statement.
Week Ending 5/5/2017
- Redesign of homepage to give more flexibility and relevance to your landing page view.
- Added autocomplete to Query to help streamline your data exploration.
- Added data side panel to Imports, Scripts, Models, and Exports to make it easier to find information about your data while you’re working.
- Added additional logging to Imports jobs to increase understanding of job performance.
- Added database names to scripts when user has access to more than one database to provide environment clarity when executing jobs.
- Added clarifying text to Python scripts parameter panel.
- Added trailing whitespace validation to server names.
- Reordered search filters to match the job types in Civis' left sidebar.
- Removed ability to create deprecated types of jobs from the Labs page.
- Fixed a bug in links from success & failure emails.
- Fixed a bug in container script credentials.
- Fixed a bug in tables linked from query.
Week Ending 4/28/2017
- Changed run history labels in scripts to reflect timestamp of each run to add more clarity to the job’s history.
- Released Client User Management.
- 3X improvement to response time for GET API actions.
- Fixed a bug in the Modeling workflow predictor column selection.
- Fixed a bug in Report Snapshot performance.
- Fixed a bug where tables with comments weren't showing up in Query.
- Fixed a bug in two-factor authentication.
Week Ending 4/21/2017
- Improved integration with Google Sheets to streamline imports and exports.
- Added links to run outputs to the Run History pane.
- Changed descriptive language from "Google Docs" to "Google Sheets".
- Upgraded Ruby to 2.3.4.
- Upgraded Bundler 1.14.6.
- Fixed a bug in SQL script notification emails.
- Fixed a bug in script runner credentials.
- Fixed a bug in Tableau where users were periodically being prompted to log in to Tableau Server.
- Fixed a bug in CASS/NCOA job connections.
- Fixed a security vulnerability to prevent a SQL injection attack.
Week Ending 4/14/2017
- Removed the ability to DELETE objects from Civis in favor of /archive endpoints.
- Added clarity to run names in the Run History pane.
- The homepage activity feed will now display one entry per import instead of multiple entries.
- Upgraded Devise gem to 4.2 and Devise-Two-Factor to 3.0. Devise is used in authentication.
- Upgraded rubyzip gem to 1.2.1.
- Fixed a bug with Tableau reports prompting users to log in to Tableau Server.
Week Ending 4/7/2017
- Improved Civis scalability by changing how Query results are stored.
- Added phone number validation to user profiles to improve 2FA SMS reliability.
- Fixed a bug with full screen report toggling.
- Fixed bugs with Google Doc integrations.
Week Ending 3/31/2017
- Switched the default Python Docker image from datascience-base to datascience-python.
- Upgraded nokogiri to v1.7.1. Nokogiri is an open-source parser used with various imports and exports.
- Fixed a bug in uploading basefiles for person matching.
- Fixed visible jobs on the /users/<username> page.
Week Ending 3/24/2017
- 6X improvement to response time on the Scripts index and associated API endpoints.
- Updated the icon for idle jobs.
- Added support for much larger job chains in a workflow.
- Improved the performance of the Imports index when filtering by Destination.
- Upgraded the Civis Python API Client to v1.4.0.
- Fixed a bug in model scoring.
- Fixed a bug in imports when dropping an empty table.
Week Ending 3/17/2017
- Tableau report snapshots now use server-side rendering to better handle more complex visualizations.
- The Last Run sorting option has been restored to all indices.
- Fixed a bug where canceled Person Match jobs were blocking subsequent match jobs.
- Fixed various bug with Google Sheet imports.
Week Ending 3/10/2017
- Updated the left side menu. Civis jobs are now grouped by functionality, which will make it easier to navigate through your work.
- In Query, the Copy Output to Clipboard function has been refactored to not use Flash.
- Background jobs created by modeling are now hidden to avoid confusion and clutter on the homepage and associated indices.
- 4X improvement to response time on the Imports, Models, and Projects indices and associated API endpoints.
- Fixed a bug in public script templates to allow templated script creation by any user.
Week Ending 3/3/2017
- Upgraded Rails to 4.2.8.
- Fixed a bug with Salesforce Import where user and client credentials weren't loading properly.
- Fixed a bug with Imports where the "If Table Exists" option of "Drop" couldn't drop an empty table.
Week Ending 2/24/2017
- Upgraded Tableau Server to 10.1
- Added more detailed error messages to Database Imports.
- Added support for non-numeric ID columns in Database Imports.
- Reduced Modeling memory consumption defaults.
- Added a helpful error message when you try to import an empty sheet in a Google Doc.
- Fixed a bug in Database Imports where users couldn't select a source database/credential they were permissioned on.
Week Ending 2/17/2017
- The API now allows you to add an object as an output of a run and then access that output via the runId.
- Fixed a bug in exports where a cloned export was redirecting to the /#/imports page.
- Fixed a bug in generating script output links.
Week Ending 2/10/2017
- Exposed the bocce_clusters API endpoint to improve visibility into Bocce jobs (such as Container Scripts).
- Disabled index sorting by last run to improve performance.
- Fixed a bug with cloning Imports and Exports.
- Fixed a bug in scoring models.
- Deprecated an unused API endpoint.
- Fixed a bug with Google Doc Imports to allow for null values in the rightmost column.
Week Ending 2/3/2017
- Tableau reports are archivable.
- Fixed a JavaScript bug in the modeling workflow.
Week Ending 1/27/2017
- Users with "view" permissions on a script can now access the run history pane.
- Added the cancel button to Imports.
- New Civis account creation links now expire after 7 days instead of 3 days to accommodate accounts created just before the weekend.
- Fixed a visual bug in Imports filters that caused the list to spill off the screen.
- Upgraded Elasticsearch to 2.0.
- Rearranged past dates on the Query page to add clarity when scrolling through the query history.
- Updated text in the Sharing pane to add clarity when a job is shared to a person hidden from the user.
Week Ending 1/20/2017
- Users have more flexible CSV settings when using the POST /scripts/sql API endpoint.
- Added the runner_user_id environment variable to the SQL Scripts API to add clarity between the user IDs of a script's author and a script's runner.
Week Ending 1/13/2017
- Allow access to credentials that have been shared with a user's non-primary groups.
- Upgrade Bundler to 1.13.6.
- Upgrade Ruby to 2.3.3.
- Additional sorting and filtering options on the /scripts/custom API endpoint.
- Users can now archive scripts with archived templates.
- Updated scripts documentation to clarify author parameter.
- Users can match an S3 file prefix with a file's owner for jobs run by the non-author.
- Restore "New Database" button to Database Import.
- Fixed a bug where users had difficulty editing notifications after a job has run.
- Fixed a bug where the notifications pane in Imports wouldn't open.
Week Ending 12/16/2016
- Improved organization and usability in Search.
- Allow “run script to update report” action when report is created in query to mimic script-backed report behavior.
- Allow resources allocated to a script template to be set at runtime using parameters to give template users more control over scripts.
- More intuitive handling of headers in CSV and Google Doc imports to reduce import errors.
- Bug fixes in report modification.
Week Ending 12/2/2016
- Search request optimization.
- Bug fix in CMO channel mapping storage.
- Bug fix to restore permissions in Tableau workbooks.
- Modeling wizard - Instead of creating a model object at the start of the wizard, collect all settings and then create and build the model when the wizard is completed.
- Remove unnecessary Redshift usage.
- Sync notification bug fixes.
- File read and validation for spot order model.
Comments
0 comments
Article is closed for comments.