airflow-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF subversion and git services (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (AIRFLOW-1774) Better handling of templated parameters in Google ML batch prediction/training operators
Date Wed, 11 Apr 2018 09:58:00 GMT

    [ https://issues.apache.org/jira/browse/AIRFLOW-1774?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16433662#comment-16433662
] 

ASF subversion and git services commented on AIRFLOW-1774:
----------------------------------------------------------

Commit 65e7025f3af378dfa825eb04d005ec1f7a422cde in incubator-airflow's branch refs/heads/master
from [~wileeam]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-airflow.git;h=65e7025 ]

[AIRFLOW-1774] Allow consistent templating of arguments in MLEngineBatchPredictionOperator

Fix a minor typo and a wrong non-default
assignment

Fix one more typo

Adapt tests to new error messages and fix another
typo

Fix exception type in utils operator test class

Improve cleansing of non-valid training and
prediciton job names

Closes #2746 from wileeam/ml-engine-prediction-
job-normalization


> Better handling of templated parameters in Google ML batch prediction/training operators
> ----------------------------------------------------------------------------------------
>
>                 Key: AIRFLOW-1774
>                 URL: https://issues.apache.org/jira/browse/AIRFLOW-1774
>             Project: Apache Airflow
>          Issue Type: Improvement
>          Components: contrib, operators
>    Affects Versions: Airflow 2.0, 1.9.0
>            Reporter: Guillermo Rodríguez Cano
>            Assignee: Guillermo Rodríguez Cano
>            Priority: Major
>
> The {{MLEngineBatchPredictionOperator}} does not support well the templated parameter
{{job_id}} due to a helper function used to detect and cleanup bad job names inhibiting the
templating engine to work. I suspect the same may happen with {{MLEngineTrainingOperator}}
as well.
> Google ML requieres a unique job id, therefore it is critical to have the possibility
to customise the job's name easily, and preferably with some data related to the DAG, for
example, appending the day to the job's name via a macro like {{ds}}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message