Standardize pymc import

[zonca]

I think it would be useful to standardize how pymc should be imported. Currently most of the examples and also the pymc3 tutorial use:

from pymc import *

Of course this should be discouraged, and it would be useful to choose a standard shortcut import.
I suggest using:

import pymc as mc

It is used by Bayesian Methods for Hackers which I guess is one of the most common ways new users start using pymc.

If there is agreement on this, it would be useful to prepare a PR with standardization of the examples and tutorial (I can do this if nobody else volunteers).

[twiecki]

pymc 2 uses import pymc as pm and so does most other pymc code I know. So if Bayesian Methods changed the imports it should all be in agreement.

[zonca]

Ok
Sorry didn't know about that.
So do you think would be useful to modify tutorial and example to that standard?

[twiecki]

I think it would be useful to change the Bayesian Methods for Hackers book to adhere to the standard :).

[zonca]

Sure, I'll ask to the author, but for pymc3 there is no standard if you are
not using it in the examples...
So if I prepare a PR on putting import pymc as pm in examples would you
merge it?
On Oct 25, 2013 7:06 AM, "Thomas Wiecki" notifications@github.com wrote:

I think it would be useful to change the Bayesian Methods for Hackers book
to adhere to the standard :).

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/367#issuecomment-27095099
.

[twiecki]

@zonca Oh I see what you mean.

I think I'm in favor of this. What do you think @jsalvatier and @fonnesbeck ?

[fonnesbeck]

Definitely, yes. Thanks for offering.

[jsalvatier]

Yes, sounds like a great idea.

On Fri, Oct 25, 2013 at 9:11 AM, Chris Fonnesbeck
notifications@github.comwrote:

Definitely, yes. Thanks for offering.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/367#issuecomment-27105132
.

[zonca]

ok, good we agreed on this, I'll prepare a PR.
However, let me advocate a little bit about my proposal:

import pymc as mc

wouldn't you consider changing the default import from pymc2 to pymc3?
afterall the API changed a lot anyways, so it wouldn't be too bad to also change the default import name.

mc as Monte Carlo is a lot more descriptive of what the package does than pm.
mc.sample, mc.Model, mc.Normal I think look a lot better than pm.sample, pm.Model, pm.Normal.

I would like also to ask also @CamDavidsonPilon's opinion on this.

[CamDavidsonPilon]

I cannot recall why I chose mc as the alias for pymc. If the community wants to standardize, I fine with changing to pm. It will stop me from writing the ugly pattern:

mcmc = mc.MCMC(...)

[twiecki]

@zonca I see your argument but disagree. I don't think the standard should change in this way because of a new major version. If anything, we might want to rename pymc to pymc3 and then do import pymc3 as pm3 which would be more consistent. Note that I'm not advocating we actually do this.

I also think we should be compatible with other packages, numpy is np, scipy is sp, statsmodels is sm etc.

[zonca]

Ok, thanks @CamDavidsonPilon @twiecki!

It looks like there is an agreement on using pm, so I am going to prepare a PR for pymc and @CamDavidsonPilon will modify Bayesian Methods for Hackers.

[twiecki]

@zonca Thanks for helping with this!