Color design paradigm in this style

[imagico]

With the recent changes/additions to area colors in dog_park (#2216, see also #2250) and playground (#2249) i have to say i can no more recognize an overall design paradigm regarding the area colors in this style.

In the last years in my eyes many improvements have been made in terms of colors here significantly enhancing the overall readability. When making changes (like the meadow/grassland and the orchard/vinyard unification) i always did so based on the idea of a common understanding of a certain overall direction in which color design is meant to steer and when others made similar changes i did sometimes disagree with the specific choices made but usually understood and concurred with the general idea. I have now the distinct impression this is no more the case and color decisions are going back to a local atomic find a free spot in color space and looks nice to me approach which had previously already led to a lot of problematic color choices.

Design is of course always a highly subjective and intuitive field so it is difficult to formulate absolute rules without exceptions. Good design tends to follow rules but also occasionally breaks them - the key however is doing so deliberately with an understanding of the effects this has and not out of ignorance or laziness.

The cartographic guidelines of this style offer relatively little substantial advise on the matter, the most significant statement is on legibility - which seems to be a relatively vague concept but in fact can be quite strong as a guiding principle, especially regarding choice of colors - if you get to a point where you do no more just determine legibility based on your personal subjective gut feeling.

I would like to ask the maintainers here to contemplate this matter. I am not asking for a recipe for good design but i ask you to consider that a design paradigm needs to be visible in the decisions made for others to successfully contribute. I currently have difficulties recognizing such a paradigm and it seems i am not the only one. This might still be an aftereffect of the move from a single maintainer to a team of several with naturally sometimes contradicting design ideas. This should however not prevent the decisions and the larger picture behind them being transparent and understandable to open minded contributors.

[kocio-pl]

As we already know, I have quite different POV than you. In a nutshell:

• I agree that legibility is "relatively vague", because it's more general rule, but for me color palette is rather a tool. It's always better to be consistent, but this rule has lower priority in my view, because its scope is limited (for example colors have to be reused in different contexts).
• I like the meadow/grassland and the orchard/vineyard unification, because it doesn't mess up with legibility - it's very good, since both rules apply.
• For a few features finding good colors is relatively easy (especially natural ones, like water, grass or sand, but also military, since red is related to danger) however for many others it's not. I don't see any color-meaning related reasons for violet borders, brown amenities, red footways, red-yellow-white roads palette, or pink medical POIs, and that is where some other rules take over, like visibility (health). For road colors - they are internally consistent, but also they are intense to be more visible.
• Leisure can be related to some less intense, nature-related colors (color design rule), be it light yellow, beige, blue or green for example, but for me being water-like is just worse (legibility rule). I don't see any of these color breaking color design rules (you suggest otherwise), but blue clearly breaks legibility rule, which has higher priority for me.

[pnorman]

In the last years in my eyes many improvements have been made in terms of colors here significantly enhancing the overall readability. When making changes (like the meadow/grassland and the orchard/vinyard unification) i always did so based on the idea of a common understanding of a certain overall direction in which color design is meant to steer and when others made similar changes i did sometimes disagree with the specific choices made but usually understood and concurred with the general idea. I have now the distinct impression this is no more the case and color decisions are going back to a local atomic find a free spot in color space and looks nice to me approach which had previously already led to a lot of problematic color choices.

I am in general agreement with this. We spent a long time cleaning up past inconsistent colour choices, we don't want to go back to the same situation.

[imagico]

Note although i mentioned the two recent color changes for dog_park and playground as cause for my problem i did not mean to blame these PRs, especially not @jdhoek for whom this was the first contribution.

This is about the big picture and overall design ideas being communicated in the guidance and decisions by the maintainers. Expressing like or dislike for design choices and making decisions based on this alone does not give me as a contributor sufficient guidance to adjust my own choices to be in line with the overall design paradigm of the project. I observe this with my own considerations but i also see it in the choices made by others.

[matthijsmelissen]

Thanks for opening this issue @imagico.

I agree with you that we're lacking a bit of a sense of direction in this respect, and that the changes we're implementing are rather ad hoc.

It's interesting to look at http://bl.ocks.org/math1985/raw/7d2b0538a8b64b9d31e2/#13.00/52.1226/5.0073, which shows the difference between the current style and the style some years ago. A question directed to everybody: if you're looking at these changes, do you still have the idea that there's a lack of direction, or can you see some consistency in the changes?

I don't think this is an issue for the maintainers only, by the way. If anybody has ideas on how to implement a more consistent design paradigm, I would be very interested. Of course, we'd also welcome PR's against CARTOGRAPHY.md.

I had a look at how other map providers are handling this. Google Maps, Bing or Mapbox Streets don't look much more consistent in color use to me. Comparing the Lch values confirms this.

Does anybody know maps with a nice and consistent colour pallet - either visually or when looking at Lch values?

[nebulon42]

Just for illustration I'm attaching two videos I have made regarding this subject (hope I haven't done that before, I don't remember).

https://vimeo.com/171819594
https://vimeo.com/171816463

[matthijsmelissen]

Thanks, cool video! Good to keep in mind that this is both data changes and mapping changes, of course.

[nebulon42]

@math1985 Thanks! But the data for each frame is the same as stated in the description

Map data from 2016-05-30 (© OpenStreetMap contributors).

Only the code of this style is changing.

[imagico]

if you're looking at these changes, do you still have the idea that there's a lack of direction, or can you see some consistency in the changes?

Note i specifically mentioned that IMO the last years overall show a good progress is consistency of color use. If you look at the history of changes you can see that there are actually not that many changes that modify or add colors. Most of these changes were not done ad hoc. But with quite a few of the more recent changes as well as color change discussion in general i see a bit of a trend away from what characterized previous discourse on color changes (see the road colors change or the wood color for good examples).

http://bl.ocks.org/math1985/raw/7d2b0538a8b64b9d31e2/#13.00/52.1226/5.0073

It is always a good idea to look back and while there are a lot of improvements in many things it is also worth noting that there are some aspects that are worse than before. Not specifically regarding colors but i especially see the situation of z13/14 as somewhat ambivalent. z11/12 are significantly better mostly due to the roads changes and z15+ are also much improved but in between there are also quite a few changes with serious downsides.

[daganzdaanda]

Looking at the before and after examples, I believe the changes in the style were overwhelmingly positive. I'm really very happy with the development up to now.
Everybody who contributed in any way - with ideas, opinions, mock-ups, tests, PRs, decisions - deserves a big Thank You!

I believe that for the future, we should continue the slow and deliberate course, not just in colour decisions but everywhere.
This doesn't mean that an "imperfect" PR can't be quickly merged sometimes, because sometimes there may not be a perfect solution, and sometimes good enough is much better than nothing. But we need to be critical about our decisions, and look at the bigger issue from time to time (like now).

(edited: was too tired to find the document that defines the "overall design paradigm" https://github.com/gravitystorm/openstreetmap-carto/blob/master/CARTOGRAPHY.md )

[kocio-pl]

@imagico

i can no more recognize an overall design paradigm regarding the area colors in this style.

It implies that you could do this earlier. My personal impression is that there was no strong "color paradigm" anyway, could you describe how did you see it back then?

[imagico]

Not easily. It is not that all color choices were made in a way i would have made them but i nearly always could see a logic behind them. In other words if choices differed from the ones i would have made this seemed to result from different weights being given to the various influencing factors but the factors considered were generally the same as for me. This seems different with some of the recent color choices. Reading up some of the old discussions regarding the road color change (here and in the user diaries) might be helpful.

I have started working on a proposal for design goals and guidelines for this style that would provide some framework for decisions here - not only color but design in general. But this will take some time.

Some good guiding principles for color selection (but certainly not the only ones) are:

• features similar in meaning and purpose for the map user should have similar colors while those different in meaning and purpose should have different colors
• small features should have stronger and heavier colors than large features
• features with distinct meaning and purpose should have stronger and heavier color than more general features.

[matthijsmelissen]

features similar in meaning and purpose for the map user should have similar colors while those different in meaning and purpose should have different colors
small features should have stronger and heavier colors than large features
features with distinct meaning and purpose should have stronger and heavier color than more general features.

I think that's already a useful first step.

[imagico]

I think that's already a useful first step.

You are aware that these three rules alone would eat the playground and dog_park colors for breakfast?

[matthijsmelissen]

You are aware that these three rules alone would eat the playground and dog_park colors for breakfast?

Yes, this convinced me of the need to reconsider these changes. Although in the playground case, I think we gained on the first point and lose on the other two points.

Concerning the pitch colour, I think even the new colour is stronger than the other landuse colours, so here I don't see as big problems.

[imagico]

Although in the playground case, I think we gained on the first point and lose on the other two points.

Seriously? To me this is the strongest point here. Among the urban open ground areas i see hardly anything further apart than playgrounds and dog parks. Note i wrote similar in meaning and purpose, not physical appearance.

Regarding pitch/track - the strongest point here would be the last one - pitch and track are highly distinct in meaning and purpose, much more than for example buildings in general (although buildings are on average smaller of course). But clearly this is a more complex matter where other aspects play a role which are not covered by these three very simple principles.

In general guidelines and principles like these should preferably be used positively to make suitable choices and not only for evaluating changes suggested for reasons outside their scope.

By the way - the term strong and heavy w.r.t. colors is a relative characterization, pure white in the context of this style for example is a very strong and heavy color.

[kocio-pl]

As for now, I really treat it like the first step and wait for the - more or less - complete set of rules to judge anything. Then I'd like to see the evaluation of all the colors we use (second step) and fine tune the rules according to this reality check (third step). Until then it's just a subjective proposal to make colors consistent, which would be nice.

By the way - we already have the general attitude toward colors documented and reality check (the final look as an important measure) is included there:

Firstly, this is a map, not merely a colourful 2-dimensional visualisation of the database. Colours should be chosen based on their effectiveness and to make things look nice, not chosen for distinctiveness.

[matthijsmelissen]

Some good guiding principles for color selection (but certainly not the only ones) are:

@imagico I would like to have this (or something along these lines) in Cartography.md. Would you like to create a pull request or shall i make one?

[imagico]

As mentioned in #2270 (comment) i am working on a more comprehensive text that includes these point among others. This will certainly require some discussion and would likely not be complete in any way either but probably better than having just these three points in isolation.

[matthijsmelissen]

Great, than I'm holding off for now!

[imagico]

With #2654 being merged now this issue is essentially to be closed as wontfix from my perspective.

Since as indicated before i have difficulties making changes in a framework based primarily on ad-hoc decisions without an overall design paradigm this is probably a good point to take some time off active development here. I am really interested to see if this new direction is going to work out and creates a useful and balanced style.

However i also think it is important to show that the current direction is not without alternatives and that a systematic development approach can solve design issues and lead to an improved map. Based on ideas i had when discussing the problems of #2654 i put together a sketch how an alternative approach to mid and low zoom design can look like. The low zoom part is not implemented since it requires more work - only for the waterbodies a practically usable solution exists so far.

I don't really want to write much about it (which would largely amount to repeating the arguments from #2654 anyway). Some explanations for what i did can be found in the commit comments. There are also a number of loose ends hanging 'in the air' so to speak that do not really harmonize well with the rest at the moment. The idea was not to get every detail right but to create a consistent and workable overall concept. From perspective of the current state in master these changes are likely more radical than #2654 in some ways. I will probably try out some other changes based on this in the future that i don't feel make much sense with the current colors and design direction here.

https://github.com/imagico/openstreetmap-carto/tree/alternative-colors

[kocio-pl]

I'm glad that you show some alternative and it looks nice. But this is very late (we were not able to compare it while testing midzoom for weeks or months) and still rough, as you say, it doesn't help with busy background and I'm also confused why do you think changing farmland color from orange to a grass-green is more intuitive than less intensive orange. I'm surprised, because you were against learning new colors and not being able to recognize landuse types, both of which are more evident with your code:

Warsaw, z10
current master

your approach

Rivers are less visible in the forest:
current master

your approach

There are also some quirks with different water colors next to each other:

[imagico]

I am not really interested in discussing this here on the level we have discussed #2654 - which has been pretty pointless and frustrating. I did not primarily want to show a different resulting map (even though i prefer the results with my changes) but a different approach to design.

I also understand that with no more detailed explanation of the specific ideas implemented and without the ability to actually see this being extended to the low zoom levels it might not be readily visible what the reasoning behind these changes is. But that is not my aim here, i already tried and failed to convince you of the principal concept of systematic design here, in #2462 and in #2654. I really want to give the new direction of this style a chance and see how it fares in terms of readability, maintainability and all the other goals we discussed. I am not convinced that my approach is better. If it turns out to be great to make decisions mostly ad hoc, popular with map users, attracting many new developers who improve the map to become better usable and better serving the diverse geography and diverse user base we have on Earth i will be the last to complain. But if not it is important to be aware of the alternatives.

[kocio-pl]

I also understand that with no more detailed explanation of the specific ideas implemented and without the ability to actually see this being extended to the low zoom levels it might not be readily visible what the reasoning behind these changes is.

For me that's the core of a problem: you tell about your systematic approach (BTW: color fading is also a system for me), yet you don't show if it's really working. So I have to believe you may eventually come with a good solution, but I have no proof for a long time.

I can wait some time, when it's not blocking or contradicting other operations, but partial implementation, which is still not ready to test your general ideas, is not convincing me. I can't even tell if you're right or wrong, because it's still the black box.

[imagico]

As i have written in #2654 (comment) i cannot make you see the points i see. As everyone can see i have tried really hard to argue for my approach, both as a general concept (here) in form of practical guidelines (in #2462) and in a specific case (in #2654).

As @pnorman likes to point out this style is fairly unique in terms of technical complexity but this also applies to design complexity. Regarding the approach to making design decisions in light of this complexity we are essentially in uncharted waters. I really hope the direction this is taking at the moment turns out to be successful in improving the map as per the goals we have but my intuition tells me otherwise (and i have learned to trust my intuition in this kind of thing). Being able to evaluate this in the future depends on known alternatives and my sketch aims to offer one.

[kocio-pl]

@nebulon42

Just for illustration I'm attaching two videos I have made regarding this subject (hope I haven't done that before, I don't remember).

https://vimeo.com/171819594
https://vimeo.com/171816463

Do you have some script how to do timelapse videos of changes in osm-carto?

Lately I have added link to your videos here:
https://wiki.openstreetmap.org/wiki/Standard_tile_layer#External_links

but it would be nice to make updated versions or just create videos of some other areas.

[nebulon42]

Unfortunately, this was done completely manually for my talk at FOSSGIS. I checked out every tag, rendered the same area with Kosmtik and saved it to a image file. My notes on how I created the video show the following:

mogrify -crop 1804x894+1+1 -gravity SouthEast *.png
ffmpeg -r 1.5 -i %d.png -c:v libx264 -profile:v high -r 25 -crf 20 -pix_fmt yuv420p output.mp4

I don't remember why the mogrify step was necessary. But the ffmpg part produces a video out of consecutive images. In theory it should be possible to automate all this. Want to give it a try? :)

[kocio-pl]

Thanks! I leave it as a pet project when I will like to try something different than typical PRs and project management issues. =}

[kocio-pl]

BTW: what's the license for these videos? I think they are interesting enough to be imported here:

https://commons.wikimedia.org/wiki/Category:OpenStreetMap_maps_of_Salzburg_(state)

so it could be visible on the OSM Wiki too.

[nebulon42]

I thought I specified the license and I did, but Vimeo does a bad job displaying the license. It is CC0 for both videos.

[matthijsmelissen]

Nice graphic of Google's color paradigm:

https://storage.googleapis.com/gweb-uniblog-publish-prod/images/Google_Maps_Color.max-1000x1000.jpg

[imagico]

I am labeling this as consensus needed because it essentially documents the core problem behind a lot of current issues/PRs like

#3843
#2353
#2691
#3891
#3670
#2905
#3607

which are all hampered by the lack of consensus on an overall color concept among the maintainers.

[SomeoneElseOSM]

@imagico based on comments at #4384 (comment) , are you effectively saying that until this issue is resolved no other pull requests to this style are possible? Along with many other people I've encouraged people who want to contribute here to do so in diary entries and elsewhere; it sounds like the people who've actually followed that advice have been wasting their time.

Do any of the other regular maintainers have a view on this?

[imagico]

Not at all - I have not made a negative review of #4384 and i have stated above here i don't think that i necessarily know better than anyone else what the best direction is for this style.

I have presented and will continue to present the reasoning behind color design choices and strategies i consider viable as well as critically comment choices i consider not to be. It is up to others what they make of this. But i also want to be very clear about one thing: If we cannot resolve this issue and come to some sort of consensus about the design principles for this style our ability to work towards the goals of this project and in particular also to recruit competent designers to work on it will be seriously hampered.