New "Turbo" colormap

[noamross]

Google's released a new "Turbo" map that is designed as a replacement for "Jet". It is designed as a compromise between viridis-type maps and the useful many-color properties of rainbow maps. It doesn't quite meet the stringent rules colors in this package meet, but it does have a coherent philosophy and some distinct advantages:

https://ai.googleblog.com/2019/08/turbo-improved-rainbow-colormap-for.html .

Should we include Turbo in viridis? Inclusion in viridis makes it rapidly available in ggplot2 and lots of other packages, which is helpful if the purpose is to serve people who want these popular maps. On the other hand it makes viridis less coherent. If so, we probably want to modify the vignette to explain the difference.

[noamross]

Some action on this already: https://gist.github.com/jlmelville/be981e2f36485d8ef9616aef60fd52ab

[sjmgarnier]

@noamross I don't mind including it but don't really have the bandwidth to work on viridis at the moment. There are still a couple of PRs pending and it's in need of being updated on CRAN. The best I can do is mid-September I'm afraid.

[noamross]

No rush at all! This can definitely go slow, just wanted to gather some input. Would be happy to do the PRs and update the vignette if there's demand.

[talgalili]

I think it would be great, but it sounds like something that is worth its own package, no?

On Mon, Aug 26, 2019, 17:48 Noam Ross notifications@github.com wrote:

No rush at all! This can definitely go slow, just wanted to gather some input. Would be happy to do the PRs and update the vignette if there's demand.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/sjmgarnier/viridis/issues/65?email_source=notifications&email_token=AAHOJBWSSXAEPUNHG5CSDLDQGPUL5A5CNFSM4IOWXMCKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5ETFRY#issuecomment-524890823, or mute the thread https://github.com/notifications/unsubscribe-auth/AAHOJBTJGREFI4A45J74OP3QGPUL5ANCNFSM4IOWXMCA .

[noamross]

@talgalili That's why I think maybe we shouldn't. On the other hand since viridis is already popular and imported by ggplot2, anything we add here anything we add here gets instant availability and audience. I would guess the ggplot2 authors don't want to add a lot of colorscale-specific packages. Perhaps one of that team (@hadley @thomasp85 @karawoo @yutannihilation @clauswilke, sorry not sure who to tag!) has a thought on this.

[thomasp85]

We've had loose discussions about using it as a basis for some of our colour scales, but nothing more... If it ends up being part of standard ggplot2 then import is something we'd have to consider, but anyone is free to create palette packages

[noamross]

Right, I meant that, if "Turbo" is included in viridis[Lite], with no changes to scales or ggplot2 you get scale_color_viridis_c(option="turbo"), just as now you also get scale_color_viridis_c(option="cividis"), which we added later.

[clauswilke]

I'd say go for it. I don't see any downside. I would recommend against starting yet another new color palette package.

[talgalili]

Your reasons makes sense, I support adding turbo to this package.

On Mon, Aug 26, 2019, 21:05 Claus Wilke notifications@github.com wrote:

I'd say go for it. I don't see any downside. I would recommend against starting yet another new color palette package.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/sjmgarnier/viridis/issues/65?email_source=notifications&email_token=AAHOJBQGD4YWRDUHCNDQUFTQGQLNBA5CNFSM4IOWXMCKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5FFH5A#issuecomment-524964852, or mute the thread https://github.com/notifications/unsubscribe-auth/AAHOJBWGJX23SO4MRAYGKFDQGQLNBANCNFSM4IOWXMCA .

[yutannihilation]

From ggplot2's side, I don't find any argument against this. It's viridis to decide what options it should have.

OTOH, as a user of viridis, I would say this is concerning.

On the other hand it makes viridis less coherent.

Matplotlib seems to give it a try. Can we wait for their conclusion? I think now viridis package can be considered as "a collection of newer color palettes in matplotlib", so, if matplotlib is to implement Turbo, I think it's natural viridis package also gets that option. https://github.com/matplotlib/matplotlib/issues/15091

[clauswilke]

I looked at the Turbo color map some more. It is clearly a diverging map, even if its creator thinks it can be used for sequential data (https://ai.googleblog.com/2019/08/turbo-improved-rainbow-colormap-for.html). As such, it should probably be implemented in a scale function that provides a midpoint argument, and neither scale_color_viridis_c() nor scale_color_viridis_d() do so. A better fit might be to place it with the diverging color scales in the colorspace package.

[clauswilke]

Actually, never mind. The chroma curve is all over the place, so you won't be able to express it in the framework of the colorspace package.

library(colorspace)
A<-c((0.18995),(0.19483),(0.19956),(0.20415),(0.2086),(0.21291),(0.21708),(0.22111),(0.225),(0.22875),(0.23236),(0.23582),(0.23915),(0.24234),(0.24539),(0.2483),(0.25107),(0.25369),(0.25618),(0.25853),(0.26074),(0.2628),(0.26473),(0.26652),(0.26816),(0.26967),(0.27103),(0.27226),(0.27334),(0.27429),(0.27509),(0.27576),(0.27628),(0.27667),(0.27691),(0.27701),(0.27698),(0.2768),(0.27648),(0.27603),(0.27543),(0.27469),(0.27381),(0.27273),(0.27106),(0.26878),(0.26592),(0.26252),(0.25862),(0.25425),(0.24946),(0.24427),(0.23874),(0.23288),(0.22676),(0.22039),(0.21382),(0.20708),(0.20021),(0.19326),(0.18625),(0.17923),(0.17223),(0.16529),(0.15844),(0.15173),(0.14519),(0.13886),(0.13278),(0.12698),(0.12151),(0.11639),(0.11167),(0.10738),(0.10357),(0.10026),(0.0975),(0.09532),(0.09377),(0.09287),(0.09267),(0.0932),(0.09451),(0.09662),(0.09958),(0.10342),(0.10815),(0.11374),(0.12014),(0.12733),(0.13526),(0.14391),(0.15323),(0.16319),(0.17377),(0.18491),(0.19659),(0.20877),(0.22142),(0.23449),(0.24797),(0.2618),(0.27597),(0.29042),(0.30513),(0.32006),(0.33517),(0.35043),(0.36581),(0.38127),(0.39678),(0.41229),(0.42778),(0.44321),(0.45854),(0.47375),(0.48879),(0.50362),(0.51822),(0.53255),(0.54658),(0.56026),(0.57357),(0.58646),(0.59891),(0.61088),(0.62233),(0.63323),(0.64362),(0.65394),(0.66428),(0.67462),(0.68494),(0.69525),(0.70553),(0.71577),(0.72596),(0.7361),(0.74617),(0.75617),(0.76608),(0.77591),(0.78563),(0.79524),(0.80473),(0.8141),(0.82333),(0.83241),(0.84133),(0.8501),(0.85868),(0.86709),(0.8753),(0.88331),(0.89112),(0.8987),(0.90605),(0.91317),(0.92004),(0.92666),(0.93301),(0.93909),(0.94489),(0.95039),(0.9556),(0.96049),(0.96507),(0.96931),(0.97323),(0.97679),(0.98),(0.98289),(0.98549),(0.98781),(0.98986),(0.99163),(0.99314),(0.99438),(0.99535),(0.99607),(0.99654),(0.99675),(0.99672),(0.99644),(0.99593),(0.99517),(0.99419),(0.99297),(0.99153),(0.98987),(0.98799),(0.9859),(0.9836),(0.98108),(0.97837),(0.97545),(0.97234),(0.96904),(0.96555),(0.96187),(0.95801),(0.95398),(0.94977),(0.94538),(0.94084),(0.93612),(0.93125),(0.92623),(0.92105),(0.91572),(0.91024),(0.90463),(0.89888),(0.89298),(0.88691),(0.88066),(0.87422),(0.8676),(0.86079),(0.8538),(0.84662),(0.83926),(0.83172),(0.82399),(0.81608),(0.80799),(0.79971),(0.79125),(0.7826),(0.77377),(0.76476),(0.75556),(0.74617),(0.73661),(0.72686),(0.71692),(0.7068),(0.6965),(0.68602),(0.67535),(0.66449),(0.65345),(0.64223),(0.63082),(0.61923),(0.60746),(0.5955),(0.58336),(0.57103),(0.55852),(0.54583),(0.53295),(0.51989),(0.50664),(0.49321),(0.4796))
B<-c((0.07176),(0.08339),(0.09498),(0.10652),(0.11802),(0.12947),(0.14087),(0.15223),(0.16354),(0.17481),(0.18603),(0.1972),(0.20833),(0.21941),(0.23044),(0.24143),(0.25237),(0.26327),(0.27412),(0.28492),(0.29568),(0.30639),(0.31706),(0.32768),(0.33825),(0.34878),(0.35926),(0.3697),(0.38008),(0.39043),(0.40072),(0.41097),(0.42118),(0.43134),(0.44145),(0.45152),(0.46153),(0.47151),(0.48144),(0.49132),(0.50115),(0.51094),(0.52069),(0.5304),(0.54015),(0.54995),(0.55979),(0.56967),(0.57958),(0.5895),(0.59943),(0.60937),(0.61931),(0.62923),(0.63913),(0.64901),(0.65886),(0.66866),(0.67842),(0.68812),(0.69775),(0.70732),(0.7168),(0.7262),(0.73551),(0.74472),(0.75381),(0.76279),(0.77165),(0.78037),(0.78896),(0.7974),(0.80569),(0.81381),(0.82177),(0.82955),(0.83714),(0.84455),(0.85175),(0.85875),(0.86554),(0.87211),(0.87844),(0.88454),(0.8904),(0.896),(0.90142),(0.90673),(0.91193),(0.91701),(0.92197),(0.9268),(0.93151),(0.93609),(0.94053),(0.94484),(0.94901),(0.95304),(0.95692),(0.96065),(0.96423),(0.96765),(0.97092),(0.97403),(0.97697),(0.97974),(0.98234),(0.98477),(0.98702),(0.98909),(0.99098),(0.99268),(0.99419),(0.99551),(0.99663),(0.99755),(0.99828),(0.99879),(0.9991),(0.99919),(0.99907),(0.99873),(0.99817),(0.99739),(0.99638),(0.99514),(0.99366),(0.99195),(0.98999),(0.98775),(0.98524),(0.98246),(0.97941),(0.9761),(0.97255),(0.96875),(0.9647),(0.96043),(0.95593),(0.95121),(0.94627),(0.94113),(0.93579),(0.93025),(0.92452),(0.91861),(0.91253),(0.90627),(0.89986),(0.89328),(0.88655),(0.87968),(0.87267),(0.86553),(0.85826),(0.85087),(0.84337),(0.83576),(0.82806),(0.82025),(0.81236),(0.80439),(0.79634),(0.78823),(0.78005),(0.77181),(0.76352),(0.75519),(0.74682),(0.73842),(0.73),(0.7214),(0.7125),(0.7033),(0.69382),(0.68408),(0.67408),(0.66386),(0.65341),(0.64277),(0.63193),(0.62093),(0.60977),(0.59846),(0.58703),(0.57549),(0.56386),(0.55214),(0.54036),(0.52854),(0.51667),(0.50479),(0.49291),(0.48104),(0.4692),(0.4574),(0.44565),(0.43399),(0.42241),(0.41093),(0.39958),(0.38836),(0.37729),(0.36638),(0.35566),(0.34513),(0.33482),(0.32473),(0.31489),(0.3053),(0.29599),(0.28696),(0.27824),(0.26981),(0.26152),(0.25334),(0.24526),(0.2373),(0.22945),(0.2217),(0.21407),(0.20654),(0.19912),(0.19182),(0.18462),(0.17753),(0.17055),(0.16368),(0.15693),(0.15028),(0.14374),(0.13731),(0.13098),(0.12477),(0.11867),(0.11268),(0.1068),(0.10102),(0.09536),(0.0898),(0.08436),(0.07902),(0.0738),(0.06868),(0.06367),(0.05878),(0.05399),(0.04931),(0.04474),(0.04028),(0.03593),(0.03169),(0.02756),(0.02354),(0.01963),(0.01583))
C<-c((0.23217),(0.26149),(0.29024),(0.31844),(0.34607),(0.37314),(0.39964),(0.42558),(0.45096),(0.47578),(0.50004),(0.52373),(0.54686),(0.56942),(0.59142),(0.61286),(0.63374),(0.65406),(0.67381),(0.693),(0.71162),(0.72968),(0.74718),(0.76412),(0.7805),(0.79631),(0.81156),(0.82624),(0.84037),(0.85393),(0.86692),(0.87936),(0.89123),(0.90254),(0.91328),(0.92347),(0.93309),(0.94214),(0.95064),(0.95857),(0.96594),(0.97275),(0.97899),(0.98461),(0.9893),(0.99303),(0.99583),(0.99773),(0.99876),(0.99896),(0.99835),(0.99697),(0.99485),(0.99202),(0.98851),(0.98436),(0.97959),(0.97423),(0.96833),(0.9619),(0.95498),(0.94761),(0.93981),(0.93161),(0.92305),(0.91416),(0.90496),(0.8955),(0.8858),(0.8759),(0.86581),(0.85559),(0.84525),(0.83484),(0.82437),(0.81389),(0.80342),(0.79299),(0.78264),(0.7724),(0.7623),(0.75237),(0.74265),(0.73316),(0.72393),(0.715),(0.70599),(0.69651),(0.6866),(0.67627),(0.66556),(0.65448),(0.64308),(0.63137),(0.61938),(0.60713),(0.59466),(0.58199),(0.56914),(0.55614),(0.54303),(0.52981),(0.51653),(0.50321),(0.48987),(0.47654),(0.46325),(0.45002),(0.43688),(0.42386),(0.41098),(0.39826),(0.38575),(0.37345),(0.3614),(0.34963),(0.33816),(0.32701),(0.31622),(0.30581),(0.29581),(0.28623),(0.27712),(0.26849),(0.26038),(0.2528),(0.24579),(0.23937),(0.23356),(0.22835),(0.2237),(0.2196),(0.21602),(0.21294),(0.21032),(0.20815),(0.2064),(0.20504),(0.20406),(0.20343),(0.20311),(0.2031),(0.20336),(0.20386),(0.20459),(0.20552),(0.20663),(0.20788),(0.20926),(0.21074),(0.2123),(0.21391),(0.21555),(0.21719),(0.2188),(0.22038),(0.22188),(0.22328),(0.22456),(0.2257),(0.22667),(0.22744),(0.228),(0.22831),(0.22836),(0.22811),(0.22754),(0.22663),(0.22536),(0.22369),(0.22161),(0.21918),(0.2165),(0.21358),(0.21043),(0.20706),(0.20348),(0.19971),(0.19577),(0.19165),(0.18738),(0.18297),(0.17842),(0.17376),(0.16899),(0.16412),(0.15918),(0.15417),(0.1491),(0.14398),(0.13883),(0.13367),(0.12849),(0.12332),(0.11817),(0.11305),(0.10797),(0.10294),(0.09798),(0.0931),(0.08831),(0.08362),(0.07905),(0.07461),(0.07031),(0.06616),(0.06218),(0.05837),(0.05475),(0.05134),(0.04814),(0.04516),(0.04243),(0.03993),(0.03753),(0.03521),(0.03297),(0.03082),(0.02875),(0.02677),(0.02487),(0.02305),(0.02131),(0.01966),(0.01809),(0.0166),(0.0152),(0.01387),(0.01264),(0.01148),(0.01041),(0.00942),(0.00851),(0.00769),(0.00695),(0.00629),(0.00571),(0.00522),(0.00481),(0.00449),(0.00424),(0.00408),(0.00401),(0.00401),(0.0041),(0.00427),(0.00453),(0.00486),(0.00529),(0.00579),(0.00638),(0.00705),(0.0078),(0.00863),(0.00955),(0.01055))

turbo_colormap_data <- cbind(A,B,C) 
turbo_colormap_data_sRGB <- sRGB(turbo_colormap_data)
turbo_colormap_data_HEX <- hex(turbo_colormap_data_sRGB)

specplot(turbo_colormap_data_HEX)

^{Created on 2019-08-26 by the reprex package (v0.3.0)}

[sjmgarnier]

As such, it should probably be implemented in a scale function that provides a midpoint argument, and neither scale_color_viridis_c() nor scale_color_viridis_d() do so

They do have a begin and end argument. Maybe not ideal but that should be enough for a savvy user to change the location of the midpoint, no?

[noamross]

Matplotlib seems to give it a try. Can we wait for their conclusion?

I agree that we should wait and see, especially as it seems Turbo may be tweaked as it goes through matplotlib review. In the meantime, I made a fork of viridisLite if you want to try it out in ggplots: remotes::install_github('noamross/viridisLite@turbo')

[yutannihilation]

As such, it should probably be implemented in a scale function that provides a midpoint argument,

FWIW, scale_*_distiller() provides diverging colour palettes without midpoint (c.f. https://ggplot2.tidyverse.org/reference/scale_brewer.html).

[yutannihilation]

It seems the next release of matplotlib will include Turbo.

https://github.com/matplotlib/matplotlib/pull/15275

[sjmgarnier]

@noamross turbo is now in viridisLite. I'll work on updating viridis accordingly very soon.

[yutannihilation]

I'm preparing pull requests on scales and ggplot2 just to reflect the new options to their docs.

• https://github.com/r-lib/scales/pull/306
• https://github.com/tidyverse/ggplot2/pull/4412

[sjmgarnier]

@yutannihilation Thanks! I submitted viridisLite to CRAN yesterday, awaiting final approval from them. Then, I will do the reverse dependence checks for viridis (can't do it until viridisLite is on CRAN) before preparing the CRAN submission. That might not be before the end of the week depending on how fast I hear back from CRAN.

[yutannihilation]

Congrats on the CRAN release! :tada:

[sjmgarnier]

@yutannihilation Thanks! It's just viridisLite for now. I need to change a few things in viridis to resolve a reverse dependence issue. Hopefully tomorrow.

[sjmgarnier]

@yutannihilation I just got confirmation that viridis is on its way to CRAN.

[yutannihilation]

Thanks. To be clear, users are already able to use new options with ggplot2 if they installed the newest version of viridisLite. The only problem is the documentation; ggplot2 refers to the documentation of scales package, and I expect the new version of scales package won't be released anytime soon, so I expect ggplot2's documentation will stay the old description for a while, unfortunately... Anyway, it's out of the scope of this issue. Congrats on the CRAN release of viridis too!

[sjmgarnier]

@yutannihilation I know. I was just letting you (and everyone else) know why I'm closing this issue.

[yutannihilation]

I see, thanks.