[
https://issues.apache.org/jira/browse/JENA-601?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13840037#comment-13840037
]
Andy Seaborne commented on JENA-601:
------------------------------------
+1
Some thing work currently, some do not because support was added after various
bits of I/O code were written. Cataloguing what works and what does not will
help pin down the areas to fix:
*ARQ*
Works:
{noformat}
SELECT *
FROM <file:D.ttl.gz>
{ ?s ?p ?o }
{noformat}
Does not work:
{noformat}
SELECT *
FROM <http://localhost/D.ttl.gz>
{ ?s ?p ?o }
{noformat}
*RIOT*
Same pattern. {{LocatorURL}} does not uncompress automagically.
{noformat}
riot D.ttl.gz
{noformat}
Parsing foo.nt.gz does work at the moment, other compression formats do not)
Ideas:
* Maybe keep {{LocatorURL}} and add one that does add a decompress wrapper.
The standard {{StreamManager}} would include the auto-decompress version.
* Alternative: Add to interface {{Locator}} an operation
{{openUncompressed(String)}}.
> Provide better support for compressed input formats
> ---------------------------------------------------
>
> Key: JENA-601
> URL: https://issues.apache.org/jira/browse/JENA-601
> Project: Apache Jena
> Issue Type: Improvement
> Components: RIOT
> Affects Versions: Jena 2.11.0
> Reporter: Rob Vesse
>
> Currently Jena has little or not support for compressed input formats. There
> are the odd cases where some consideration is given e.g.
> - {{RDFLanguages.filenameToLang()}} strips off {{.gz}} extensions to help it
> correctly detect file types
> - HTTP responses can deal with compressed responses by virtue of Apache
> HttpClient
> What would be nice is to have a better strategy for handling compressed
> inputs. For example having a registry of known compression extensions e.g.
> {{.gz}}, {{.bz2}}, {{.deflate}} which ARQ would strip off when trying to
> deduce format from the filename.
> It would also be useful if the various locator implementations took
> compression into account when opening input streams as I'm fairly sure if you
> asked ARQ to open a {{foo.nt.gz}} file it would just open a raw input stream
> and then the reading would fail.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
