[VELOCITY-714] Change ResourceLoader API to return Reader instances instead of InputStreams

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Invalid
• Component/s: Engine
• Assignee: Unassigned
• Reporter: Jarkko Viinamäki
• Created: 02/May/09 14:43
• Updated: 04/May/09 14:10
• Resolved: 04/May/09 13:27

Description

Abstract class org.apache.velocity.runtime.resource.loader.ResourceLoader contains method

public abstract InputStream getResourceStream( String source )
throws ResourceNotFoundException;

However, in org.apache.velocity.Template.process() there is code (abbreviated):

try

{ is = resourceLoader.getResourceStream(name); }

catch( ResourceNotFoundException rnfe )

{ errorCondition = rnfe; throw rnfe; }

if (is != null)
{
try
{
BufferedReader br = new BufferedReader( new InputStreamReader( is, encoding ) );
data = rsvc.parse( br, name);
...

This seems a bit silly. ResourceLoaders should obviously return Reader instances since those would allow loaders to take content encoding into account. Currently this is problematic especially with DataSourceResourceLoader (VELOCITY-599). If we want to support UTF-8 in DataSourceResourceLoader, with current API we would first have to use ResultSet.getCharacterStream to get a Reader, then read all stuff into a String, then convert it into UTF-8 encoding with getBytes call but then in the Template the whole stuff is then converted back to a Reader. That's a lot of waste.

I think it would make more sense if the ResourceLoader API contained a method:

public abstract Reader getResource( String source )
throws ResourceNotFoundException;

Issue Links

• is related to: VELOCITY-599 DataSourceResourceLoader doesn't support UTF8
• is related to: VELOCITY-451 Do some investigation into UTF-16 and InputStream / Reader interfaces