Btw, only due to this post I looked into layouts and eventually switched to KOY. I'm even starting to work on improving Arne's layout-generator! 🙈

Hi! Yes, I have already calculated a layout that should be way better for German thanthe current one. (#144 (comment))

That being said, there are two problems.

1. It's still not the final german layout as I haven't had the abillity to test it.
2. 8Vim was put on hold due to real-life things the maintainer needs to tackle. This is why it might take a while until the German layout gets implemented.

This means that if you want to start using 8vim in the near future, you'll probably have to go for the english version.

(Except for if you posess the skills to add the english layout yourself. In that case tell me, ich will es auch endlich verwenden!)

Measurements

I'm still a bit confused by the method you used. Wouldn't that method systematically score bigrams as worse, that have a bad second-letter -> first-letter flow (and score those better, that have a good second-letter -> first-letter flow)?
If I understand correctly, it seems to me that this method measures both L1 -> L2 as well as L2 -> L1, since both those values impact the measurement.

For example: (I'm using the left, rotated image you posted a couple of comments ago).
We test the letter-combination n -> r, which in my opinion is really nice.

Now if I understand your method correctly, we try to write this bigram as often as possible within the span of 10s.Thus, we write

r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> …

Later on, we test the letter-combination r -> n, which in my opinion is uncomfortable.

n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> n -> r -> …

Looking at what we wrote, both those measurements should return roughly the same number of loops (unless confirmation bias affects our performance). This, to me, seems like it is a big problem, looking at the fact that we got the same results for two wildly different movements.

The postion of the "q" is a matter of taste and makes no relevant difference in this score. That's what I meant. All positions are within the range of your high ranked 5 German layouts.

If we were to move it, I'd move it to the bottom-left spot you proposed, due to the (imo) good flow of q -> u

No, a little different. I measured the following at that time:
r -> n pause r -> n pause r -> n pause ....
resp.
n -> r pause n -> r pause n -> r pause .....

By pause I mean to lift the finger briefly and then start writing again.

I wanted to use it to make an estimate of how long it takes on average to write a given position (regardless of transitions) and use the result as an additional evaluation criterion.

In the end, however, it remains a purely subjective measurement.

The pausing was a good (and important) add-on.

How good are the best layouts I have found compared to your best ones, if we use your rating-system for all of them?

These are the writing times I got out as an average for the individual positions (as a beginner).

timing as measured (right-handed)	-->	averaged left-right differences

And my layouts were always better than the original 8pen layout (which was my goal), but unfortunately never as good as your last ones.
The score is to be understood as a mere addition, in which the transitions are not taken into account. I used it mainly as feedback for the optimization process, and now to optimize the rotation of the layout.

Which of these layouts would you choose as German standard layout (in which order)?

The one which works well with English text, so it might be important to not just use German bigrams, but English as well, maybe (if easily possible) with a slighty lower weight? Ideally the German layout would be mostly identical to the "perfect" English one, so someone familiar with one can easily learn the other.

1. I agree that we need to factor in English writeability
2. There's quite a few problems with the goal of creating layouts similar to "the perfect" English layout.
   1. Creating similar layouts is not part of the optimizer's functionality
   2. We don't know what the perfect English layout looks like
   3. It might not be possible to create proper German layouts that have a close resemblance of the perfect English layout

Well, a layout optimized for the german language must have the goal to show a good flow for german texts in the first place. In my opinion, the flow in other languages can only be an additional criterion for deciding among comparably good German layout variants.

@ThomasAH, of course, you can think about layouts that are optimized for several languages. But that is not the goal of a German layout. To be honest, i had this thought myself and experimented with it, but i gave it up when i saw that this always leads to a cutback in the flows in the individual languages.
Besides, maybe this can be solved better once we have a simple switching function (gesture) between languages.

Besides, maybe this can be solved better once we have a simple switching function (gesture) between languages.

I agree with most of the things said, but I must inject that I think learning multiple layouts isn't feasible. It's hard enough to get good at one of them.

I agree with most of the things said, but I must inject that I think learning multiple layouts isn't feasible. It's hard enough to get good at one of them.

This is the important part. It is hard enough to switch between QWERTY and QWERTZ. Maybe include the absolutely most common English words (20? 50? 100?) in the calculation to not disturb the German word flow too much?

Here is my reasoning for favoring kjotom's layout 1 out of the 3 German layouts:

1. German bigrams

• it is only slightly below the two high scores (ORIGINAL_SCORE_LIST and KJOETOM_SCORE_LISTE_NEW) of the other two German layouts
• this also applies to the total writing time of all bigrams and the number of forced stops

2. English bigrams

• it achieves in the first four lines even the best values among all 4 compared layouts, surprisingly even better than the original English 8vim layout

3. French bigrams

• also here it achieves the best values for both scores and the total time for writing all bigrams

For me it is close between kjotom's layout 1 and 2. I think both would be good, and I probably prefer 1, but I may be biased because you presented it this well :)

It might be nice to compare the total time for writing things that are typically written on smartphones. Is there a text corpus for emails, tweets and other short messages in German/English/French?

1. Good job working on criteria
2. Am I correct in assuming comparing the two scores that are furthest up (ORIGINAL_SCORE_LIST and KJOETOM_SCORE_LIST) is most important?
3. When optimizing these new layouts, how much %weight did you assign to German/English/French?
4. For future graphs, orange might be a better color than white 😉

ad 2. --> Since no one can actually decide which score is closest to reality, I simply plotted the results of the two scores (ORIGINAL_SCORE_LIST; KJOETOM_SCORE_LIST_NEW) one below the other. There is no rating in the sense of importance. I just wanted to provide a direct comparison.

ad 3. --> Since this is about German layouts, I have weighted both kjoetom layouts with 100% German (as far as I know, this also applies to Glitchy's original best layout). The English and French sections are only meant to show how well the respective German layout is suited for the English and French bigram sets.

ad 4. --> I am not satisfied with the color scheme myself either. Maybe I should not color the English 8vim layout at all, because it is displayed out of competition, so to speak. Also, maybe the scores should be more clearly distinguishable from the other evaluation criteria.

I also set up another kjoetom score (KJOETOM_SCORE_LIST_NEW_2) that not only rates moves with forced stops or/and layer 4 positions as "bad", but also those where the rotation direction has to be changed 2x between positions. So in the next overview I might also include this additional score and a layout based on it.

@ThomasAH
Not that I am aware of any such text corpora. Moreover, I would probably not prefer them to a corpus that is as comprehensive as possible. However, it would be interesting to see if the frequency of bigrams deviates to such an extent that a strongly different layout would result.

I've written some more (but short) German and English text, and the newest layout feels very good!

One thing I noticed on the number/symbol pad: The first page has the ₹ (Rupee) sign, to reach the € (Euro) sign I have to switch to the second page. Can this be changed via the layout file or should I open a feature request to make the primary currency symbol configurable (or auto-detectable via device locale) in 8vim?

Edit: I know that I can enter € via a gesture, but in my test text I had € only in conjunction with numbers.

I'm glad that you like the layout and that you get along well with it.

I can't influence the number pad via the definition file for the layout. This is implemented in 8vim itself.

But you might know that you can write "€" and "¢" directly (without switching to the number pad), as if it were a diacritical character - actually the first diacritical character at the "e" and the "c" respectively. But probably this is what you meant when you wrote "enter € via a gesture".

I can't influence the number pad via the definition file for the layout. This is implemented in 8vim itself.

I'll open a separate feature request.

But you might know that you can write "€" and "¢" directly (without switching to the number pad), as if it were a diacritical character - actually the first diacritical character at the "e" and the "c" respectively. But probably this is what you meant when you wrote "enter € via a gesture".

As I added in my "Edit" above (which probably didn't generate a new email notification):
"I know that I can enter € via a gesture, but in my test text I had € only in conjunction with numbers."
Example text: 123.45€ * 6 = 740.70€

Edit: I didn't notice your edit about seeing my edit ;-)

Here is the feature request: issues #279 (Change currency symbol on first page of number/symbol pad)

but we would have to ask @doak to implement this in his layout generator that I've used.

@kjoetom, although I am not very much involved these times, I am happy to help out here. Just let me know what you need (and give me some time).

I love that visualization!

Me too, thank you ;-)

I felt the need for a clearer presentation when I realized that the look-up table in the previous comment helps in finding the movement sequences, but not really in remembering them.

A picture is worth a thousand words. And the look up is also much more efficient and easier (at least as far as the characters are linked to letters/signs of the x-keypad).

The main reason from my point of view is that I sometimes use words or even names from another language (in my case e.g. French, Italian or Croatian) and then want to write them correctly.

On the language specific x-keypad (German in my case) these letters are not available at all. So if I need one of these letters I would have to change to another x-keypad or alternatively even another input method.

I tried to sort the letters so that the language-specific ones appear first and the others are then arranged based on the frequency in other languages.

I hope I could make this somehow comprehensible.

Anyway, I am really grateful that you made it possible to customize the x-keypads and thus try out new ideas/layouts.

Thanks for your great work.

kjoe

Hello @MrCustomizer,

As far as I know, there is currently no automatic conversion of the layout from XML to YAML format.

However, you can more or less easily get your desired YAML layout, depending on what you meant by "I don't want to train myself with the english layout which is currently provided in it".

1. If you simply want to have the layout in YAML format, as shown here in the article above, that also was contained in the XML, you can use the file from martin-braun. He has added a few extras and offers this YAML file here for download.

2. If you want a layout that is more like the German layout that comes with 8vim, you can use my layout calculator version 2.0 and build the layout yourself (here). This spreadsheet is made for Libreoffice Calc (does not work in Excel) and will give you a working YAML configuration file at the end.

I hope this helps you.
kjoe

Hello kjoe,

thanks a lot for the fast response. I've downloaded martin-braun's file. That solved my problem.

I probably haven't been clear enough on my intentions, in part because I'm new to the project and it isn't yet clear to me which layouts exist. What I meant is that the German layout provided in the app doesn't seem to be optimized for German words as it seems to be the same layout as the English version, just with umlauts on the second layer (the letter 'y' for example is on the first position, whereas in German it should be placed less prominent).

As training yourself a new layout and keyboard is a lot of effort, I didn't want to start with this layout, just to retrain myself later on with a layout which is more optimized to German words (I'm typing mostly in German on my smartphone). I'm starting to learn martin-braun's layout now, which seems pretty good for the German language to me. Thanks to martin-braun and the original creator of this layout, too. :)

I don't think you have any disadvantages with this layout compared to the original english 8vim layout, even if you write in English instead of German. At least according to my measurements (with english bigrams), the layout optimised for German and English seems to perform better than the original 8vim layout.

From this point of view, one might consider integrating it into 8vim as a 2nd default layout for English and German. ;-)
Of course this was not meant seriously. I know that many who have been using the original 8vim layout for a long time will not change. But someone new to 8vim ...