Dear all,
Thank you very much for your valuable feedback!
I'll explain a bit what I'm doing just to clarify, sorry if this spam to some.
I want to build a model for species assemblages based on
co-occurrence of taxa within an arbitrary area. I'm building a 2D
lattice in which for each cell I'm collapsing the data into a
taxonomic tree (the occurrences). For doing this I need first to
obtain the data from the gbif api and later, based on the ids (or
names) of each taxonomic level (from kingdom to occurrence) build
a tree coupled to each cell.
The implementation is done with postgresql (postgis) for storing
the raw gbif data and neo4j for storing the relation
"Being a member of the [ specie, genus, family,,,] [name/id]" The
idea is to include data from different sources similar to the
project Matthew and Jennifer had mentioned (which I'm very
interested and like to hear more) and traverse the network looking
for significant merged information.
One of the immediate problems I've found is to import big chunks
of the gbif data into my specification. Thanks to this thread I've
found the tools that are the most used by the community
(pygbif,rgbif, and python-dwca-reader). I was using urlib2 and
things like that.
I'll be happy to share any code or ideas with the people interested.
Btw, I've checked the tinkerpop project which uses the Gremlin
traversal language as independent from the DBMS.
Perhaps it's possible to use it with spark and Guoda as well?
Does GOuda is working now?
Best wishes
Juan.
Jorrit pointed out this thread to us at iDigBio. Downloading and importing data into a relational database will work great, especially if as Jan said you can cut the data size down to a reasonable amount.
Another approach we've been working on in a collaboration called GUODA [1] is to build an Apache Spark environment with pre-formatted data frames with common data sets in them for researchers to use. This approach would offer a remote service where you could write arbitrary Spark code, probably in Jupyter notebooks, to iterate over data. Spark does a lot of cool stuff including GraphX which might be of interest. This is definitely pre-alpha at this point and if anyone is interested, I'd like to hear your thoughts. I'll also be at SPNHC talking about this.
One thing we've found in working on this is that importing data into a structured data format isn't always easy. If you only want a few columns, it'll be fine. But getting the data typing, format standardization, and column name syntax of the whole width of an iDigBio record right requires some code. I looked to see if EcoData Retriever [2] had a GBIF data source and they have an eBird one that perhaps you might find useful as a starting point if you wanted to try to use someone else's code to download and import data.
For other data structures like BHL, we're kind of making stuff up since we're packaging a relational structure and not something nearly as flat as GBIF and DWC stuff.
[1] http://guoda.bio/
[2] http://www.ecodataretriever.org/
Matthew Collins
Technical Operations Manager
Advanced Computing and Information Systems Lab, ECE
University of Florida
352-392-5414
From: jorrit poelen <jhpoelen@xs4all.nl>
Sent: Monday, May 30, 2016 11:16 AM
To: Collins, Matthew; Thompson, Alexander M; Hammock, Jennifer
Subject: Fwd: [API-users] Is there any NEO4J or graph-based driver for this API ?Hey y’all:
Interesting request below on the GBIF mailing list - sounds like a perfect fit for the GUODA use cases.
Would it be too early to jump onto this thread and share our efforts/vision?
thx,
-jorrit
Begin forwarded message:
From: Jan Legind <jlegind@gbif.org>
Subject: Re: [API-users] Is there any NEO4J or graph-based driver for this API ?
Date: May 30, 2016 at 5:48:51 AM PDT
To: Mauro Cavalcanti <maurobio@gmail.com>, "Juan M. Escamilla Molgora" <j.escamillamolgora@lancaster.ac.uk>
_______________________________________________Dear Juan,Unfortunately we have no tool for creating these kind of SQL like queries to the portal. I am sure you are aware that the filters in the occurrence search pages can be applied in combination in numerous ways. The API can go even further in this regard[1], but it not well suited for retrieving occurrence records since there is a 200.000 records ceiling making it unfit for species exceeding this number.There is going be updates to the pygbif package[2] in the near future that will enable you to launch user downloads programmatically where a whole list of different species can be used as a query parameter as well as adding polygons.[3]In the meantime, Mauro’s suggestion is excellent. If you can narrow your search down until it returns a manageable download (say less than 100 million records), importing this into a database should be doable. From there, you can refine using SQL queries.Best,Jan K. Legind, GBIF Data managerFrom: API-users [mailto:api-users-bounces@lists.gbif.org] On Behalf Of Mauro Cavalcanti
Sent: 30. maj 2016 14:06
To: Juan M. Escamilla Molgora
Cc: api-users@lists.gbif.org
Subject: Re: [API-users] Is there any NEO4J or graph-based driver for this API ?Hi,
One solution I have successfully adopted for this is to download the records (either "manually" via browser or, yet better, using a Python script using the fine pygbif library), storing them into a MySQL or SQLite database and then perform the relational queries. I can provide examples if you are interested.
Best regards,2016-05-30 8:59 GMT-03:00 Juan M. Escamilla Molgora <j.escamillamolgora@lancaster.ac.uk>:Hola,
Is there any API for making relational queries like taxonomy, location or timestamp?
Thank you and best wishes
Juan
_______________________________________________
API-users mailing list
API-users@lists.gbif.org
http://lists.gbif.org/mailman/listinfo/api-users
--
API-users mailing list
API-users@lists.gbif.org
http://lists.gbif.org/mailman/listinfo/api-users
_______________________________________________ API-users mailing list API-users@lists.gbif.org http://lists.gbif.org/mailman/listinfo/api-users