[ https://issues.apache.org/jira/browse/NIFI-3402?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16240587#comment-16240587 ]
ASF GitHub Bot commented on NIFI-3402: -------------------------------------- Github user m-hogue commented on a diff in the pull request: https://github.com/apache/nifi/pull/2150#discussion_r149147685 --- Diff: nifi-nar-bundles/nifi-standard-bundle/nifi-standard-processors/src/main/java/org/apache/nifi/processors/standard/InvokeHTTP.java --- @@ -1093,6 +1133,19 @@ private Charset getCharsetFromMediaType(MediaType contentType) { return contentType != null ? contentType.charset(StandardCharsets.UTF_8) : StandardCharsets.UTF_8; } + /** + * Retrieve the directory in which OkHttp should cache responses. This method opts + * to use a temp directory to write the cache, which means that the cache will be written + * to a new location each time this processor is scheduled. + * + * Ref: https://github.com/square/okhttp/wiki/Recipes#response-caching + * + * @return the directory in which the ETag cache should be written + */ + private static File getETagCacheDir() { + return Files.createTempDir(); --- End diff -- As it's currently written, this should be done via OS-controlled temporary folder cleanup [1]. I elected to do this to 1) avoid manual cleanup and 2) avoid introducing something error prone. As this is only a cache, its durability isn't important. [1] https://google.github.io/guava/releases/19.0/api/docs/com/google/common/io/Files.html#createTempDir() > Add ETag Support to InvokeHTTP > ------------------------------ > > Key: NIFI-3402 > URL: https://issues.apache.org/jira/browse/NIFI-3402 > Project: Apache NiFi > Issue Type: Improvement > Reporter: Brandon DeVries > Assignee: Michael Hogue > Priority: Trivial > > Unlike GetHTTP, when running in "source" mode InvokeHTTP doesn't support > ETags. It will pull from a URL as often as it is scheduled to run. When > running with an input relationship, it would potentially make sense to not > use the ETag. But at least in "source" mode, it seems like it should at > least be an option. > To maintain backwards compatibility and support the non-"source" usage, I'd > suggest creating a new "Use ETag" property that defaults to false... -- This message was sent by Atlassian JIRA (v6.4.14#64029)