[ 
https://issues.apache.org/jira/browse/CAMEL-27299?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18148223#comment-18148223
 ] 

Claus Ibsen commented on CAMEL-27299:
-------------------------------------

Fixed in https://github.com/apache/camel/pull/30503 with a narrower scope than 
proposed. The binary data type transformer itself stays as it is: in 
transformDataType, inputType/outputType and action Kamelets it still reads a 
plain InputStream into a byte[], because the steps after it got a byte[] 
before, and a stream that can be read only once would be empty for the second 
reader (review of #30453, BinaryDataTypeOneShotStreamTest). Only 
KameletProducer, when it applies the binary output type of a source Kamelet, 
keeps an InputStream body and sets Content-Type. That is what the consuming 
route got before CAMEL-27258.

_Claude Code on behalf of Claus Ibsen_

> camel-core - the binary data type keeps an InputStream body, and does not 
> read it into a byte array
> ---------------------------------------------------------------------------------------------------
>
>                 Key: CAMEL-27299
>                 URL: https://issues.apache.org/jira/browse/CAMEL-27299
>             Project: Camel
>          Issue Type: Improvement
>          Components: camel-core
>            Reporter: Claus Ibsen
>            Assignee: Claus Ibsen
>            Priority: Major
>             Fix For: 4.24.0
>
>
> {{ByteArrayDataTypeTransformer}} ({{application-octet-stream}}, alias 
> {{binary}}) keeps a {{StreamCache}}, {{File}}, {{Path}} or {{WrappedFile}} 
> body (CAMEL-27258, CAMEL-27267), but reads any other body into a {{byte[]}}, 
> also a plain {{InputStream}}.
> Since CAMEL-27258, the output data type of a source Kamelet is applied before 
> the consuming route. Many source Kamelets of the catalog declare {{binary}} 
> (aws-s3-source, google-storage-source, azure-storage-blob-source, 
> ibm-cos-source and more). With stream caching turned off, their component 
> gives a plain {{InputStream}}, and the {{binary}} type now reads the whole 
> object into memory, where before the route got the stream. For large objects 
> that can exhaust the heap. It also stops the catalog from declaring 
> {{binary}} on ftp-, ftps- and sftp-source, whose large-file modes 
> ({{streamDownload}}) give a plain {{InputStream}}.
> Proposal: {{ByteArrayDataTypeTransformer}} also keeps an {{InputStream}} body 
> as it is, and only sets {{Content-Type: application/octet-stream}}. A plain 
> {{InputStream}} can be read only once, but that is what the route got before 
> the output type was applied, so nothing changes for it; a route that needs 
> bytes converts the body, as it did before.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to