Move irrational constants to its own module #23427

fredrikekre · 2017-08-24T15:00:59Z

No description provided.

giordano · 2017-08-24T16:53:05Z

base/irrationals.jl

-log(::Irrational{:e}, x::Number) = log(x)
-
-# align along = for nice Array printing
-function alignment(io::IO, x::Irrational)


Was this method removed on purpose?

No, should have been moved.

Keno · 2017-08-24T18:10:13Z

This was discussed in triage and the preferred resolution was the following:

Delete eu
Make ℯ (\mscre) a unicode alias for e (i.e. rebase Deprecate Euler number e in favor of ℯ (U+212F) #10612)
Export π and ℯ from base by default, put the ascii versions as well as the other math constants in its own module (this PR)
Doesn't all have to be done in this PR, but I wanted to note it down somewhere relevant.

simonbyrne · 2017-08-24T20:17:19Z

I think we should probably export pi as well.

Keno · 2017-08-24T20:17:56Z

seems fine as well.

fredrikekre · 2017-08-24T23:21:16Z

base/linalg/lapack.jl

@@ -5299,7 +5299,7 @@ for (bdsdc, elty) in
                u, &ldu, vt, &ldvt,
                q, iq, work, iwork, info)
            chklapackerror(info[])
-            d, e, u, vt, q, iq
+            d, e_, u, vt, q, iq


Apparently we returned that mathematical constant e from LAPACK.bdsdc!. 🤘

Just checked and e_ is just garbage after the call to xbdsdc so maybe just delete it (and adjust svd)

fredrikekre · 2017-08-24T23:38:50Z

At time of writing I think this PR now implements the new behaviour. I think the easiest would be to incorporate #10612 in this PR for the deprecation of eu etc. What should we name the new module? Constants, MathConsts, MathConstants, MathematicalConstants or other suggestions?

giordano · 2017-08-24T23:44:54Z

Note that ISO mandates that mathematical constants are to be typeset with roman (upright) font, thus using ℯ for the Euler constant is in contrast with this rule. Of course, Julia doesn't have to follow ISO on this.

simonbyrne · 2017-08-24T23:54:15Z

Irrationals, since they're all irrational?

fredrikekre · 2017-08-25T11:03:16Z

Irrationals, since they're all irrational?

In that case we should probably wrap everything regarding irrationals in that module, not just the constants.

andyferris · 2017-08-25T11:27:06Z

I actually like that we export ASCII e (and definitely pi)

KristofferC · 2017-08-25T11:33:29Z

Exporting a one letter variable is way too error prone for a general purpose language, in my opinion. I have thrown e multiple times, I have had bugs in my code due to missing initializing my own \gamma etc. The cost of someone wanting to use e as Eulers constant having to do an import or using the Unicode variable seems very worthwhile if it avoids putting a one letter variable into all code.

andyferris · 2017-08-25T11:41:15Z

Doesn't the exact same argument apply to I? (Are there other one letter ASCII exports?)

KristofferC · 2017-08-25T11:48:18Z

It sort of does, although I guess it is more rare to use I (capital letter) in general code than e. Anyway, I think the path forward here is already decided and most the arguments for and against have been made, so perhaps not necessary to rehash them.

SimonDanisch · 2017-08-25T11:57:12Z

I have never used e but use I quite a bit and definitely run into problems with it being exported ;) Guess it all depends on the naming style of a particular person!

andyferris · 2017-08-25T11:59:33Z

Anyway, I think the path forward here is already decided and most the arguments for and against have been made, so perhaps not necessary to rehash them.

Is this referring to:

This was discussed in triage

? (edit: or is there a link to something that could have been added to the OP?) (edit2: I guess the discussion was in #10612...)

fredrikekre · 2017-08-25T12:02:02Z

I would be the last single ASCII export after this. We could make that 𝐈 (\mbfI) or something, but thats a separate discussion.

giordano · 2017-08-25T13:37:57Z

This is a breaking change, because the type of the Euler's number is different. This can probably be handled in Compat.jl by defining the same constant.

ararslan · 2017-08-26T03:26:23Z

I must say I still don't like the Unicode alias for e. It's too visually similar to ASCII e which just makes it confusing. I think we should either keep exporting e or put it in the math constants module with no exported alias.

fredrikekre · 2017-08-28T18:36:40Z

Could use some guidance on how to deprecate stuff here. Are we OK with just make this breaking? I tried some things e.g.

export e
e = Irrational{:e}()
Float64(::Irrational{:e}) = (depwarn("e is moved etc...", :Float64); Float64(ℯ))
# ...

but then you get a conflict when using Base.MathConstants:

WARNING: using MathConstants.e in module Main conflicts with an existing identifier.

Any way to actually deprecate this?

ararslan · 2017-08-28T18:38:07Z

Should be able to use @deprecate_binding e MathConstants.e

fredrikekre · 2017-08-28T18:44:04Z

Hmmmm, I was sure I tried that, but seems to work. Does not give a very helpful message though:

julia> e
WARNING: Base.e is deprecated.
  likely near no file:0
ℯ = 2.7182818284590...

PallHaraldsson · 2017-08-28T20:48:42Z

@Keno "Make ℯ (\mscre)" What does "mscre" stand for anyway? I'm never going to remember this unless I know..

Can we have (also) \e and/or \euler for "Euler's number"?: https://en.wikipedia.org/wiki/e (mathematical constant).

Just in case it's not a good idea, note there's also: https://en.wikipedia.org/wiki/Euler–Mascheroni constant

fredrikekre · 2017-08-28T20:50:51Z

\mscre = math-script-e

Keno · 2017-08-28T20:52:08Z

mscre stands for "math script e". We can and should add another unicode alias. \e has been suggested, but it's annoying, because there's a bunch of other completions that start with \e..

- only export pi, π and ℯ from Base - remove eu as alias for ℯ - add \euler as tab completion for ℯ

fredrikekre · 2017-08-28T21:23:15Z

Ok, so lets decide on a name for the module?
This is also a problem:

julia> e
WARNING: Base.e is deprecated.
  likely near no file:0
ℯ = 2.7182818284590...

julia> e
WARNING: Base.e is deprecated.
  likely near no file:0
ℯ = 2.7182818284590...

The warning is rather cryptic, and does not suggest the alternative and also prints more than once. This works though;

julia> using Base.MathConstants

julia> e
ℯ = 2.7182818284590...

Any ideas would be much appreciated.

StefanKarpinski · 2017-08-31T16:41:02Z

NEWS.md

@@ -363,6 +363,15 @@ Deprecated or removed

  * `diagm(A::BitMatrix)` has been deprecated, use `diagm(vec(A))` instead ([#23373]).

+  * `ℯ` (written as `\mscre<TAB>` or `\euler<TAB>`) is the new default for Euler's


What does it mean for this to be "the new default"? Default in what sense?

The one being exported, and also typeof(e) has changed from Irrational{:e} to Irrational{:ℯ}.

Ok, might be good to clarify that if you're willing!

Yes, I think my own wording is kinda weird too.

giordano reviewed Aug 24, 2017

View reviewed changes

Keno added this to the 1.0 milestone Aug 24, 2017

fredrikekre force-pushed the fe/constants branch from e64f1b5 to c309521 Compare August 24, 2017 23:16

fredrikekre commented Aug 24, 2017

View reviewed changes

fredrikekre added deprecation This change introduces or involves a deprecation needs news A NEWS entry is required for this change labels Aug 24, 2017

move mathematical constants to Base.MathConstants

abc6aa2

- only export pi, π and ℯ from Base - remove eu as alias for ℯ - add \euler as tab completion for ℯ

fredrikekre force-pushed the fe/constants branch from c309521 to abc6aa2 Compare August 28, 2017 21:14

fredrikekre removed the needs news A NEWS entry is required for this change label Aug 28, 2017

StefanKarpinski assigned fredrikekre Aug 31, 2017

StefanKarpinski approved these changes Aug 31, 2017

View reviewed changes

StefanKarpinski merged commit 8f6f5aa into master Aug 31, 2017

StefanKarpinski deleted the fe/constants branch August 31, 2017 16:46

StefanKarpinski mentioned this pull request Aug 31, 2017

better deprecation messages for math constants #23526

Closed

This was referenced Aug 31, 2017

clarification about ℯ #23534

Merged

Deprecate Euler number e in favor of ℯ (U+212F) #10612

Closed

fredrikekre mentioned this pull request Jan 8, 2018

update syntax JuliaEditorSupport/Julia-sublime#69

Merged

cormullion mentioned this pull request Mar 18, 2018

NEWS.md is getting a bit untidy #26508

Closed

ararslan mentioned this pull request Mar 18, 2019

Renaming Catalan's constant so Catalan numbers can be defined more na… #12706

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Move irrational constants to its own module #23427

Move irrational constants to its own module #23427

fredrikekre commented Aug 24, 2017

giordano Aug 24, 2017

fredrikekre Aug 24, 2017

Keno commented Aug 24, 2017

simonbyrne commented Aug 24, 2017

Keno commented Aug 24, 2017

fredrikekre Aug 24, 2017

andreasnoack Aug 25, 2017

fredrikekre commented Aug 24, 2017

giordano commented Aug 24, 2017

simonbyrne commented Aug 24, 2017

fredrikekre commented Aug 25, 2017 •

edited

Loading

andyferris commented Aug 25, 2017

KristofferC commented Aug 25, 2017 •

edited

Loading

andyferris commented Aug 25, 2017

KristofferC commented Aug 25, 2017

SimonDanisch commented Aug 25, 2017

andyferris commented Aug 25, 2017 •

edited

Loading

fredrikekre commented Aug 25, 2017 •

edited

Loading

giordano commented Aug 25, 2017

ararslan commented Aug 26, 2017

fredrikekre commented Aug 28, 2017

ararslan commented Aug 28, 2017

fredrikekre commented Aug 28, 2017

PallHaraldsson commented Aug 28, 2017 •

edited

Loading

fredrikekre commented Aug 28, 2017

Keno commented Aug 28, 2017

fredrikekre commented Aug 28, 2017

StefanKarpinski Aug 31, 2017

fredrikekre Aug 31, 2017

StefanKarpinski Aug 31, 2017

fredrikekre Aug 31, 2017

		@@ -363,6 +363,15 @@ Deprecated or removed

		* `diagm(A::BitMatrix)` has been deprecated, use `diagm(vec(A))` instead ([#23373]).

		* `ℯ` (written as `\mscre<TAB>` or `\euler<TAB>`) is the new default for Euler's

Move irrational constants to its own module #23427

Move irrational constants to its own module #23427

Conversation

fredrikekre commented Aug 24, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Keno commented Aug 24, 2017

simonbyrne commented Aug 24, 2017

Keno commented Aug 24, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fredrikekre commented Aug 24, 2017

giordano commented Aug 24, 2017

simonbyrne commented Aug 24, 2017

fredrikekre commented Aug 25, 2017 • edited Loading

andyferris commented Aug 25, 2017

KristofferC commented Aug 25, 2017 • edited Loading

andyferris commented Aug 25, 2017

KristofferC commented Aug 25, 2017

SimonDanisch commented Aug 25, 2017

andyferris commented Aug 25, 2017 • edited Loading

fredrikekre commented Aug 25, 2017 • edited Loading

giordano commented Aug 25, 2017

ararslan commented Aug 26, 2017

fredrikekre commented Aug 28, 2017

ararslan commented Aug 28, 2017

fredrikekre commented Aug 28, 2017

PallHaraldsson commented Aug 28, 2017 • edited Loading

fredrikekre commented Aug 28, 2017

Keno commented Aug 28, 2017

fredrikekre commented Aug 28, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fredrikekre commented Aug 25, 2017 •

edited

Loading

KristofferC commented Aug 25, 2017 •

edited

Loading

andyferris commented Aug 25, 2017 •

edited

Loading

fredrikekre commented Aug 25, 2017 •

edited

Loading

PallHaraldsson commented Aug 28, 2017 •

edited

Loading