[
https://issues.apache.org/jira/browse/JENA-1740?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16914324#comment-16914324
]
Timo Homburg commented on JENA-1740:
------------------------------------
Hello Greg,
I agree that the vector data functions could be worth adding. I think it is a
question of strategy. I could release a package containing only the vector
functions and then have an extended raster package. Would you think this is a
good idea?
At the same time, I think I can contribute some changes to your geosparql
extension which I did in my code.
For raster data: In my view the options are like this:
# "Convert the raster pixel-by-pixel band-by-band to RDF" - INSANE. Too many
triples
# Vectorization of the raster: Not optimal as the vectorization might not fit
your needs in term of raster size etc.
# Integration of the raster image in the triple store as a literal type, a
link or a binary blob (RasterWKB - The postgis way) - I would prefer this
implementation as it is very flexible, typically people only need a ROI (region
of interest) of the raster and often metadata of raster images is rather
queried than the image itself. Concerning performance: An interesting question
would be in which way the performance of processing of raster deviates from
POSTGIS. If federated queries are used and the whole raster needs to be loaded
over the internet I would imagine the bottleneck is the download time. Apart
from that I currently see no difference concerning the processing speed of the
raster processing functions as compared to POSTGIS. (Both may contain rasters
in WKB, rasters can be indexed for better performance and annotated, bounding
boxes can be extracted to exclude the raster loading for some queries) But I
would be interested to know if I am wrong at this assumption.
# Lastly: Should the raster be stored in the triple store? I do not know, it
could be linked or encoded in a literal. I speculate it would depend on the
implementation of the triple store if this has an impact. Also interested to
get other opinions on this.
Timo
> Extending Apache Jena with more geospatial functions
> ----------------------------------------------------
>
> Key: JENA-1740
> URL: https://issues.apache.org/jira/browse/JENA-1740
> Project: Apache Jena
> Issue Type: Improvement
> Reporter: Timo Homburg
> Priority: Minor
>
> Hello,
> I am the developer of a project which I named postgis-jena
> (https://github.com/i3mainz/postgis-jena), originally aimed at extending
> JenaARQ with query functions common in databases such as POSTGIS but now
> going beyond that approach and also integrating support for raster data
> (still WIP). I am building up work being done by Greg Albistons extension
> geosparql-jena.
> Is this extension of interest for Apache Jena as a plugin or in another
> context?
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
