egg|nomz|egg changed the topic of #kspacademia to: https://gist.github.com/pdn4kd/164b9b85435d87afbec0c3a7e69d3e6d | Dogs are cats. Spiders are cat interferometers. | Космизм сегодня! | Document well, for tomorrow you may get mauled by a ネコバス. | <UmbralRaptor> egg|nomz|egg: generally if your eyes are dewing over, that's not the weather. | <ferram4> I shall beat my problems to death with an engineer.
<egg>
bofh: I'm also tempted to make a pun on "character" and ask about U+A66E CYRILLIC LETTER MULTIOCULAR O,
FluffyFoxeh has quit [Read error: Connection reset by peer]
FluffyFoxeh has joined #kspacademia
<UmbralRaptor>
"We also neglect the transverse Doppler redshift predicted by the special relativity theory (lower than 1 m/s and identical for αCen and Proxima)."
* UmbralRaptor
amused.
<bofh>
rofl
<bofh>
less than 1m/s? that seems oddly low for SR transverse doppler redshift.
<bofh>
could be misremembering tho.
<egg|cell|egg>
!Wpn bofh
* Qboid
gives bofh a surface-mount asm.js which vaguely resembles a prime
<kmath>
<JCTArtStudio> Self portrait with #dromaeosaur. Demonstrating correct handling techniques for a tame individual. #SafetyFirst… https://t.co/HtEjrAi5bp
<kmath>
<JCTArtStudio> Self portrait with #dromaeosaur. Demonstrating correct handling techniques for a tame individual. #SafetyFirst… https://t.co/HtEjrAi5bp
<UmbralRaptor>
>_>
<egg>
!wpn bofh
* Qboid
gives bofh a francium cube
<egg>
!wpn UmbralRaptor
* Qboid
gives UmbralRaptor a polarized stellarator
<UmbralRaptor>
!wpn bofh && egg
* Qboid
gives bofh && egg a fish-compatible lamp which vaguely resembles a blurglecruncheon
<whitequark>
!wpn egg
* Qboid
gives egg a perpendicular creole
<UmbralRaptor>
orthonornalization as part of language design.
tawny- has joined #kspacademia
awang has quit [Read error: Connection reset by peer]
<Qboid>
egg: Seems that Wolfram is unable to understand that.
<iximeow>
!waluigi
<SnoopJeDi>
[distant WAAAAHHHHHH]
<UmbralRaptor>
egg: CDT (UT-5)
<egg>
!wa time UTC-5
<Qboid>
egg: current time in UTC -5 hours: 12:37:49 pm GMT-5 | Sunday, April 1, 2018
<egg>
... what is 12 pm
<SnoopJeDi>
afternoon
<egg>
so noon, rather than midnight?
<SnoopJeDi>
right, midnight is 12:00 am. ante meridiem and post meridiem
<SnoopJeDi>
even though that's still ambiguous for both of those, heh. I guess the idea is that 12:01 pm is clearly afternoon, and there's nothing particularly special about 12:00->12:01 so make 12:00 pm
<egg>
yeah I know what these acronyms expand to, but whether noon is before or after noon is endlessly mindboggling to me
<egg>
SnoopJeDi: yeah, but then you get inconsistent discontinuities between am/pm and 12/1, hence confusion
<SnoopJeDi>
it's neither and you can blame whoever took a bad language and added it to dates :P
<SnoopJeDi>
egg, ? it's perfectly consistent, the point itself is on a transition and the label goes with the thing you're transitioning into
<SnoopJeDi>
it's just less than ideal because "ideally" noon/midnight would be neither and have their own suffix (been tried and failed)
<egg>
(if you want to be sure nobody is going to have the stuff ready at the right time and date in an international context, use a deadline of 12 a.m. on 1/12)
<SnoopJeDi>
oh, you meant as a date
<egg>
SnoopJeDi: I mean that the bit "a.m./p.m." changes at 12:00, whereas the discontinuity (12:59 > 1:00) happens at 13:00
<SnoopJeDi>
no datetime system alleviates the "know your audience" criterion of good communication :)
<egg>
SnoopJeDi: some exacerbate it though
* whitequark
stares at egg
<SnoopJeDi>
egg, the discontinuity in what? the fact that it's modulo-12?
<whitequark>
oh god I never realized that
<egg>
SnoopJeDi: yes
<whitequark>
that's horrible
<egg>
SnoopJeDi: it's modulo 12 offset 1
<iximeow>
egg: thankfully most of my professors used 11:59pm as deadlines rather than 12:00
<iximeow>
informed, no doubt, by semesters of students that thought they meant the following day
<egg>
SnoopJeDi: so the number of fractional hours is in the range [1, 13[, with a discontinuity at that endpoint
<egg>
SnoopJeDi: and the am/pm bit flips at 12, which is inside the interval
<egg>
that's *really* confusing because you may expect the bit flip to be at the discontinuity
Technicalfool_ has joined #kspacademia
<egg>
SnoopJeDi: that's unique to the a.m./p.m. bit, all the other usual parts of a date and time have discontinuities that align
Technicalfool has quit [Ping timeout: 383 seconds]
<egg>
!wpn whitequark
* Qboid
gives whitequark a minimum quagma 🐍
<UmbralRaptor>
Why is there a snek in the collider?
pizzaoverhead has joined #kspacademia
<UmbralRaptor>
And yes, am/pm is horrible.
<UmbralRaptor>
Naturally anyone who uses 24 hour times to alleviate ambiguity is mocked. >_<
<whitequark>
mocked?
<whitequark>
I mock anyone who uses am/pm
<egg>
!wpn whitequark
* Qboid
gives whitequark an entangled conifer
<egg>
!wpn UmbralRaptor
* Qboid
gives UmbralRaptor a Kahan ‽
<UmbralRaptor>
Eggcelent.
<whitequark>
!wpn egg
* Qboid
gives egg a nasal cuneiform resonator
<whitequark>
that
<whitequark>
sounds disturbingly existential
<egg>
whitequark: bofh: UmbralRaptor: extremely horrifying idea: we could make constexpr two-line-element literals
<egg>
(I mean evidently it fits in two pages even when generalized to n other than -2 :-p)
<egg>
ferram4: so interestingly the best magic number (that minimizes the error after the integer arithmetic) isn't the best magic number after one round of Newton (or after one round of Halley, or several rounds, etc.)
<egg>
with Halley and Newton it's still easy to find the magic numbers minimizing the error after k iterations though
<ferram4>
This actually doesn't surprise me too much.
<egg>
with higher order Householder methods it seems trickier because they can introduce new error extrema
<egg>
(that's a Kepler orbit with Pluto's eccentricity)
<egg>
bofh: semimajor axis is 1 m, unit gravitational parameter, so the absolute error in energy is basically a relative error
<egg>
bofh: below 1 % Newton-Delambre-Størmer-Verlet-Leapfrog no longer has the best error
<egg>
maybe 0,1 %
<bofh>
Interesting. Yeah, not surprised.
<bofh>
wtf is up with Quinlan-Tremaine order 8 tho? It seems like it's a *massive* improvement to order 10, and then very minor past that, in terms of convergence rate.
<bofh>
(also WTF why does one of the lines have a cusp inflection point in it?)
<egg>
bofh: extremely WTF at this cusp too
<egg>
not a clue
<egg>
I think it's the SRKN6B by Blanes and Moan?
<bofh>
Yeah, that's what I see as well based on colour mapping.
<bofh>
Kinda curious, should re-test it and see what happens around ~2800 evaluations or so.
<egg>
bofh: I'm not sure I see what you mean with the Quinlans
<egg>
oh you mean Quinlan-Tremaine order 8, vs. the Quinlans of order 8?
<egg>
yeah well 9 years happened that's what happened :-p
<egg>
they found out about stability/resonance issues I think? anyway if you're looking for order 8 the quinlan-tremaine is essentially useless
<bofh>
OH. Yeah, except it seems Quinlan1999Order8B is better than QuinlanTremaine1990Order8 but *worse* than QuinlanTremaine1990Order10 (unless I'm misreading)
<egg>
well, it's lower-order
<bofh>
Point.
<egg>
but basically the quinlan tremaine are "use with great care after analysis"
<egg>
they can get really wonky with resonances
<egg>
we use them for the solar system because we tested the error and it was best with Quinlan Tremaine order 12
<bofh>
Oh, they're symmetric linear schemes instead of symplectic RK schemes.
<bofh>
wtfffffffffffff
<egg>
but by default we go with a Blanes-Moan, too risky to go with the multisteps blindly
<bofh>
I would not expect those to work better ever, huh.
<egg>
bofh: the multisteps are massively more efficient
<egg>
because if something isn't really integrable at this stepsize for an SPRK it will switch to integrating a different system and be happy
<bofh>
yep, I remember that issue from not too long ago :P basically Pol/Bop encounters massively amplified the error & the state corruption persisted and got worse each time.
<egg>
if the step is too large for a multistep method, the state is corrupt and it's downhill from there
<egg>
bofh: that's fun though, it keeps tracking the correct trajectory
<egg>
+ oscillations
<egg>
I think I saw some papers that ran elements from a multistep integration through a filter to drop things with absurdly high frequencies and got something useful out of it
<bofh>
okay, that's interesting. and yeah, if it kept tracking the correct trajectory but would acquire oscillations then yeah, LPFing the integrator state when it suddenly gained a lot of high-freq components would work quite well & I want to test this myself now.
<SnoopJeDi>
egg, re: earlier conversation, yes I agree dates are a Hard Problem :P
<egg>
SnoopJeDi: it's a regular occurence at work that somebody says something is on 9/10 and others think "oh good we have a month to prepare the slides" when "tomorrow" was meant
<egg>
k/n where k and n are in [1, 12] is an evil date format
<SnoopJeDi>
egg, yea I just say [Month] [Day] in communications/docs, and give times in the recipient's timezone
<egg>
yeah spelling out the month or ISO 8601 extended format calendar generally obviate confusion
<SnoopJeDi>
I think enough people I talk to could manage to screw up YYYYMMDD tbqh
<egg>
basic format is confusing
<egg>
YYYY-MM-DD is OK for my porpoises
<egg>
also I guess UmbralRaptor would want JD
<SnoopJeDi>
yea it's definitely a matter of preference in my case
<UmbralRaptor>
import astropy
* egg
pokes bofh in the rootns
<SnoopJeDi>
not without paying the tariff, you lil' science scamp
<UmbralRaptor>
(JD is really annoying in size if you aren't going astronomy during eg: the Old Kingdom, or for Gilgamesh,
<UmbralRaptor>
)
<SnoopJeDi>
`cal 9 1752` sums up 100% of how I feel about calendars
<UmbralRaptor>
?
<egg>
SnoopJeDi: interestingly Principia will disagree with that, it assumes Gregorian from 1582 (and refuses to parse older dates because ISO 8601 mandates proleptic Gregorian and we're not doing Maya long count conversions)
<SnoopJeDi>
egg, interesting, although refusal seems like good design for the level of precision you use internally
<SnoopJeDi>
(AFAICT)
<egg>
SnoopJeDi: we use a double from J2000
<egg>
but we could go with a double-double, astropy style
<egg>
we actually have a DoublePrecision<Instant> type and it's used quite a bit with the integrators
<SnoopJeDi>
mmm, double-double
<UmbralRaptor>
SnoopJeDi: ah. Yeah, calendars are annoying.
* egg
prods bofh with a rootn
* SnoopJeDi
glances at Antarctic timezones
<UmbralRaptor>
Also, 5 Guys > In-n-Out
<SnoopJeDi>
yea.
<SnoopJeDi>
UmbralRaptor, agreed
<UmbralRaptor>
For that matter, Burger King > In-n-Out
<SnoopJeDi>
damn, 🔥
<UmbralRaptor>
(reason: both are a reasonable drive. In-n-Out would require flying >3000 km)
<SnoopJeDi>
It wasn't bad but In-n-Out seems like mostly Californians hyping California
<bofh>
egg: wtf, it mandates proleptic *Gregorian* instead of Julian? why in the fuck.
<SnoopJeDi>
otoh, we have Whataburger here so...Texans are equally kool-aidy
<egg>
bofh: it requires agreement between the parties in order to interchange pre-1582 dates (because those dates are silly), but if they are used they're proleptic gregorian
<UmbralRaptor>
Whataburger exists in more than like 2 cities, though.
<SnoopJeDi>
there's an In-n-Out in Waco
<SnoopJeDi>
...and that's about all there is in Waco :P
<bofh>
egg: yes but *why*? Julian makes infinitely more sense.
<UmbralRaptor>
There's a Waco, CA?
<SnoopJeDi>
no just...for some reason they wanted to put a location in kinda-nowhere TX
<SnoopJeDi>
not being entirely fair to Waco but still...
<UmbralRaptor>
(Also, if you mean TX, does Janet Reno grill the burgers?)
<SnoopJeDi>
not as far as I could tell
<egg>
bofh: because (1) scope: "This International Standard is applicable whenever representation of dates in the Gregorian calendar, times in the 24-hour timekeeping system, time intervals and recurring time intervals or of the formats of these representations are included in information interchange."
<egg>
NOTE In the proleptic Gregorian calendar, the calendar year [0000] is a leap year.
<UmbralRaptor>
I have questions.
<egg>
bofh: basically if you wanted to support Julian (and Julian proleptic) you'd need *some sort* of addition to the markup to make it explicit that you're using a different calendar (especially with the post-1582 mess, but also with people who use proleptic gregorian etc.). That's out of scope for 8601, so it just defines the one calendar
<UmbralRaptor>
Namely, how did you get a 0-indexed calendar?
<bofh>
who the hell actually uses proleptic Gregorian for pre-1852 dates?
pizzaoverhead has quit [Read error: Connection reset by peer]
<egg>
bofh: mayan scholars
<egg>
because they don't have to deal with Julian nonsense :-p
<SnoopJeDi>
I thought that was a reply to zero-indexing at first
<SnoopJeDi>
!wpn -add:adj zero-based
<Qboid>
SnoopJeDi: Adjective added!
<egg>
UmbralRaptor: by defining your calendar maximally simply from Gregorian rules :-)
<UmbralRaptor>
Principia feature that would be a fun prank: display dates using Maya long count.
<egg>
I mean I need to look at date display tbh
<egg>
we don't display dates in principia, but we do dump them in logs or something
<egg>
and right now they're shown as seconds from J2000
<egg>
that is an annoying date format
<egg>
I have conversion from ISO 8601 (at compile time!) I might as well have the reverse
<bofh>
seconds from J2000 isn't the worst date format tbh.
<egg>
bofh: I mean that's not a contest I want to enter
<egg>
bofh: here's a worst date format: 8601 calendar date basic format interpreted as an integer and printed as hex
<egg>
the date is 133EDB2 (egg time)
<bofh>
so like, days and remainder seconds from J2000 honestly is a decent simple date format.
<bofh>
ROFL
<egg>
bofh: possibly worse: same but with 8601 ordinal date basic format: 1ECB2C
<egg>
no on second thought the calendar date as int to hex is probably the worst of the two
<egg>
bofh: although they're monotone so that's too easy
<bofh>
egg: right, let me readthru Approximate_rootn.pdf and then do some benchmarking, oops.
* egg
pets bofh
* egg
purrs
<bofh>
ordinal dates are honestly gross to begin with tbh.
* bofh
meows
<egg>
Felger liked them (haven't seen him in a while though)
<UmbralRaptor>
egg: worst would convert mixed endian.
<egg>
UmbralRaptor: good point
<egg>
but back when he was in charge the release dates of RealSolarSystem were ordinal dates :D