7
Blaze: the future of numpy? David Fischer April 20, 2013 David Fischer San Diego Python April 20, 2013 1/7

Blaze the-evolution-of-numpy

Embed Size (px)

Citation preview

Page 1: Blaze the-evolution-of-numpy

Blaze: the future of numpy?

David Fischer

April 20, 2013

David Fischer San Diego Python April 20, 2013 1 / 7

Page 2: Blaze the-evolution-of-numpy

What is Blaze?

http://blaze.pydata.org/

extends but does not replace numpy

backed by continuum.io

$3M in DARPA funding

David Fischer San Diego Python April 20, 2013 2 / 7

Page 3: Blaze the-evolution-of-numpy

What do I need blaze for?

# What about this?

arange(15 * 10**9).reshape(150, 10**8)

David Fischer San Diego Python April 20, 2013 3 / 7

Page 4: Blaze the-evolution-of-numpy

Features

Operate on data on disk similarly to data in memory

Custom shapes (not just N*M)

Custom types

Improved persistence formats

Add metadata to arrays

David Fischer San Diego Python April 20, 2013 4 / 7

Page 5: Blaze the-evolution-of-numpy

Is it ready?

Installation is a pain – Anaconda helps

“Pre-alpha” software

Documentation is forthcoming

Using blaze means committing to blaze development

1.0 release targeted for 2014

David Fischer San Diego Python April 20, 2013 5 / 7

Page 6: Blaze the-evolution-of-numpy

Future features

Database features like SQL & indexes

(improved) multi-core, multi-processor, distributed

(easier) GPU integration

David Fischer San Diego Python April 20, 2013 6 / 7

Page 7: Blaze the-evolution-of-numpy

Resources

http://blaze.pydata.org/

http://continuum.io/blog/blaze

https://github.com/ContinuumIO/blaze-core

http://www.slideshare.net/teoliphant/blaze-a-largescale-arrayoriented-infrastructure-for-python

https://speakerdeck.com/sdiehl/blaze-next-generation-numpy

http://pythonsd.org

David Fischer San Diego Python April 20, 2013 7 / 7