R/generated_client.R)CivisFuture class: removed deprecated earlySignal argument (issue #263)CivisFuture class: removed deprecated envir field from future::Future() and updated remote eval to use globals instead (issue #262)R/generated_client.R)with_mock() function. See issue #253.resolved() function (previously imported from the future package) and
changed its input argument from x to future, to stay consistent with the resolved.CivisFuture()
functionfuture > 1.31.0, "succeeded" is no longer recognized as a valid state for a
future. When the Civis API returns a "success" or "succeeded" status, we now map this
to a "finished" state for future objects.local argument to the CivisFuture() function. The local argument
is defunct for future versions > 1.31.0.R/generated_client.R)fetch_until will now fetch results from the API until the result is empty (i.e. returns
NULL) instead of FALSE. This change prevents a coercion error of trying to boolean compare
a NULL value in R. (#243)latest rather than 3 (#238).
This could require users to update code as it will automatically be updated to the latest version of
R deployed on Civis Platform moving forward (including R 4.0.0) (#238)enhancements_[verb]_cass_ncoa, exports_[verb]_files,
and imports_[verb]_files have updated documentation as a result. (#240)stringsAsFactors=FALSE default. (#232).database parameter in read_civis now accepts character
objects of class glue.query_civisand query_civis_file now respect credential and
credential_id args respectively when passed through ... (#219)run_template takes an additional value of JSONValue. If TRUE,
function returns the JSON output instead of the file id.errorDescription (#214)civis.default_polling_interval to globally
set the interval for polling via await (#204).write_civis_file now uploads data frames as CSVs by default.read_civis.numeric now has read.csv as the
default argument for using.DBI, dplyr and dbplyr from dependencies.feather, and ggplot2 (used for CivisML features)
are now suggested rather than imported.source. roxygen2 is now suggested rather
than imported..until argument of fetch_until must now be only a function rather
than a function or a formula.stringr, purrr, and devtools are removed from dependencies.await_all is now vectorized over two arguments (typically run and job
ids) as intended (#171).civis_platform are now tagged to civisanalytics/datascience-r:2 for
compatibility with R 3.6.0.IO
write_civis_file.data.frame uploads a data frame to S3 as a CSV.Utitilies for platform scripts
civis_script annotates an integer as a valid custom, container, or other platform script.read_civis.civis_script is added to easily read script outputs into memory.fetch_output_file_ids returns file ids of run outputs for a civis_script.fetch_output returns the output of scripts_list_*_runs_outputs for a civis_script.write_job_output for adding File run outputs to scripts.run_civis can be used to run an arbitrary R exporession in a Civis Platform containerrun_template can be used to run a template job, and return file IDs of results.write_civis.numeric now correctly syncs with headers [#150].write_civis gains header, credential_id and import_args arguments to more
flexibly import data ([#150, #149, #130]).coef.civis_ml which returns civis_ml model coefficients in the format of stats::coefcoef.civis_mlget_feature_importance to civis_ml_utilsget_feature_importance to test_civis_ml_utilsfeature_importances.rds to tests/testthat/datacivis_ml. Fixes #131.tryCatch when generating the client. Fixes #138.write_civis_file automatically retries on upload failures.library(civis) now works even if the API key has expired or no API key is present.read_civis now correctly reads exports from Redshift containing null values.read_civis and download_civis.Local dataframes can now be sent to CivisML as feather files.
Print CivisML training warnings when model results are fetched from Platform.
predict.civis_ml and create_and_run_pred gain a dvs_to_predict argument which allows users to restrict output predictions to a subset of targets from a multi-output model.
query_civis_file exports a "schema.tablename", sql("query"), or existing sql script id to S3 and returns the file id.
write_civis gains a diststyle argument for controlling the distribution of tables on Redshift.
write_civis_file now defaults to the file path for the name argument instead of requiring it.read_civis.numeric, write_civis.numeric, download_civis.numeric, and query_civis.numeric now fail with a better error message when called with x = NA of type numeric.hidden argument of write_civis now works.resolved.CivisFuture now gets the status of the platform job directly and updates the state of the future.civis_ml documentation with the following CivisML 2.1
features and defaults:
n_jobs is dynamically calculated by default for training as
well as prediction.feather-format is now allowed as an input format.value.CivisFuture now prints error logs from failed jobs, throws a catcheable
error, and fetches the job logs automatically.civis_ml_fetch_existing can handle runs with missing
metrics.json files.civis_ml will use the most recent CivisML version available, falling back to CivisML 1.1.write_civis calls imports_post_files correctly after a change in the argument order.publish_rmd and publish_htmlpolling_interval can now be specified in civis_mlcall_api only retries on 413, 429, 502, 503, 504 as in civis-python.write_civis_file uploads objects larger than 50MB
in multiple parts, and supports objects up to 5TB.
Uploads occur in parallel if a future::plan has been specified.write_civis_file now don't expire by default..Rmd and .html files as Civis Platform reports.write_civis.numeric is provided to sync a CSV file on S3 to Redshift given a file id.plan(civis_platform) has been added to evaluate R expressions on Civis Platform
using the future API.multilayer_perceptron_regressor, multilayer_perceptron_classifier,
stacking_regressor, and stacking_classifiercross_validation_parameters = "hyperband" in gradient_boosting_classifier, random_forest_classifier, extra_trees_classifier, multilayer_perceptron_classifier, gradient_boosting_regressor, random_forest_regressor, extra_trees_regressor, and multilayer_perceptron_regressor.n_jobs > 1 (default 4)validation_data = "skip"."cpu_requested", "memory_requested", and "disk_requested"
arguments in predict.civis_ml.configure and configure.win now use R_HOME to find Rscript instead of assuming that Rscript is on the path.configure uses bin/sh instead of bin/bashconfigure.win uses Rscript.exe for windows buildsconfigure.win does not use bin/bash.civis_ml objects have been improved.fetch_predictions is provided to read results from a prediction job into memory from S3.civis_ml_regressor shows y-yhat binned histogram using ggplot2.civis_ml_classifier shows the decile plot using ggplot2.civis_ml objects provide histograms of the OOS scores using ggplot2.CIVIS_ML_CLASSIFIERS and CIVIS_ML_REGRESSORS.fetch_logs now works for any civis_api object with a valid *_runs_logs method, and any civis_error object.expires_at parameter of write_civis_file has been clarified.civis_ml now throws an appropriate error message if a model type is called with multiple dependent variables but doesn't support them.civis_ml, with the CivisML error logs being printed to the console.empty_result_error is now thrown from read_civis.sql if the query returns no results.as_function deprecation from purrr replaced.get_default_database uses that database.read_civis now reads from a temporary file when the file size is > than 2^31-1 bytes.civis_ml wrapper for CivisML 1.1.civis_ml_sparse_logistic, civis_ml_gradient_boosting_classifier, civis_ml_random_forest_classifier, civis_ml_extra_trees_classifier, civis_ml_sparse_linear_regressor, civis_ml_sparse_ridge_regressor,
civis_ml_gradient_boosting_regressor, civis_ml_random_forest_regressor, and
civis_ml_extra_trees_regressor.civis_table, civis_file used to indicate data sources for civis_mlpredict.civis_ml runs a CivisML scoring job on platform for a given model and data.fetch_logs can be used to fetch logs from a model run.fetch_oos_scores can be used to return out of sample/fold scores from a model.civis_ml_fetch_existing returns an existing platform job.read_civis reads a"schema.table", a sql("query"), or a file id to a data frame.read_civis sql queries are read only.write_civis uploads a data frame or local csv to Redshift.write_civis_file uploads a serialized R object or an unserialized local file to the files endpoint (S3).query_civis runs an arbitrary "query", sql("query"), or a query id.query_civis only returns query meta data.download_civis downloads a"schema.table", a sql("query"), or a file id to a file.download_civis can export in parallel using split = TRUE.options(civis.default_db = "my_database").refresh_table can be used to refresh table meta datatransfer_table can be used to transfer tables between redshift databasesget_table_id can be used to retrieve a table iddbWriteTable, dbSentStatement, dbReadTable, dbListTablespublish_rmd, publish_html return the report id.publish_rmd.print.civis_api method for API responses hides attributes.await and await_all are provided for polling job/task completion.await and await_all use an exponential backoff retry time for polling if .interval is not set (the default).await and await_all throw a civis_await_error on platform errors, and print the platform error message along with job and run ids.get_error can be used to obtain additional debugging data from civis_await_error (platform errors).await and await_all throw a civis_timeout_error if .timeout was specified.NULL can be passed directly to the API.scripts_post_sqlLICENCE is now BSD-3.CODE_OF_CONDUCT.mdCHANGELOG.md