On the other hand, I use Extend-Esc for CapsLock now.
]]>Widely Alternating A (comma-preserving):
digraphs Frequency Period
c,z 0.00% 124201
c,s 0.18% 563
c,x 0.03% 3775
s,z 0.00% 182657
s,x 0.00% 1221420
x,z 0.00% 11520819
l,q 0.00% 273970
l,n 0.07% 1436
n,q 0.01% 16622
m,t 0.03% 3591
m,w 0.00% 59742
m,v 0.00% 248412
b,m 0.09% 1071
b,t 0.02% 5089
b,d 0.01% 19058
b,w 0.00% 73719
b,g 0.00% 133329
b,v 0.00% 25430
t,w 0.09% 1125
t,v 0.00% 72322
d,m 0.02% 5306
d,t 0.00% 23857
d,w 0.01% 8547
d,g 0.03% 2927
d,v 0.02% 5168
g,m 0.01% 9023
g,t 0.02% 5756
g,w 0.00% 134189
g,v 0.00% 679006
v,w 0.00% 2492385
j,y 0.00% 1676819
j,p 0.00% 315391
j,k 0.00% 541068
p,y 0.04% 2726
f,j 0.00% 628376
f,y 0.01% 10151
f,p 0.00% 59631
f,h 0.00% 41213
f,k 0.00% 58678
h,j 0.00% 584478
h,y 0.05% 1976
h,p 0.09% 1053
h,k 0.00% 29247
k,y 0.01% 15925
k,p 0.00% 64683
e,o 0.11% 900
i,u 0.12% 843
Total 1.08% 92
Unsurprisingly, custom layouts can beat generic ones like Colemak for nonstandard touchtyping techniques. In the case of colemak, hitting C with the index finger increases same-finger by 0.31% above the usual 1.07%:
Colemak (C typed with index finger)
digraphs Frequency Period
q,z 0.00% 3088227
a,q 0.00% 41722
a,z 0.04% 2719
w,x 0.00% 437621
r,w 0.04% 2298
r,x 0.00% 83181
f,s 0.02% 4410
c,p 0.00% 40152
c,t 0.49% 205
c,g 0.00% 89710
c,d 0.00% 21052
c,v 0.00% 226656
p,t 0.11% 908
p,v 0.00% 240337
t,v 0.00% 72322
g,p 0.00% 112932
g,t 0.02% 5756
g,v 0.00% 679006
d,p 0.00% 34114
d,t 0.00% 23857
d,g 0.03% 2927
d,v 0.02% 5168
b,c 0.00% 39697
b,p 0.00% 53884
b,t 0.02% 5089
b,g 0.00% 133329
b,d 0.01% 19058
b,v 0.00% 25430
k,l 0.03% 3302
k,n 0.10% 970
k,m 0.00% 51826
h,k 0.00% 29247
h,j 0.00% 584478
h,l 0.01% 7037
h,n 0.04% 2729
h,m 0.01% 7532
j,k 0.00% 541068
j,l 0.00% 443193
j,n 0.01% 8902
j,m 0.00% 428334
l,n 0.07% 1436
l,m 0.03% 3620
m,n 0.04% 2737
e,u 0.18% 560
i,y 0.03% 3443
Total 1.38% 72
As a side curiosity, if Colemak f and c were swapped, that would bring same-finger almost back down to the original and put the more common c at the arguably more appropriate "almost home row" QWERTY E position, at the cost of moved Ctrl-C:
Colemak (F,C swapped ; F typed with index finger)
digraphs Frequency Period
q,z 0.00% 3088227
a,q 0.00% 41722
a,z 0.04% 2719
w,x 0.00% 437621
r,w 0.04% 2298
r,x 0.00% 83181
c,s 0.18% 563
f,p 0.00% 59631
f,t 0.09% 1145
f,g 0.00% 58273
f,v 0.00% 1195982
p,t 0.11% 908
p,v 0.00% 240337
t,v 0.00% 72322
g,p 0.00% 112932
g,t 0.02% 5756
g,v 0.00% 679006
d,f 0.00% 30544
d,p 0.00% 34114
d,t 0.00% 23857
d,g 0.03% 2927
d,v 0.02% 5168
b,f 0.00% 238156
b,p 0.00% 53884
b,t 0.02% 5089
b,g 0.00% 133329
b,d 0.01% 19058
b,v 0.00% 25430
k,l 0.03% 3302
k,n 0.10% 970
k,m 0.00% 51826
h,k 0.00% 29247
h,j 0.00% 584478
h,l 0.01% 7037
h,n 0.04% 2729
h,m 0.01% 7532
j,k 0.00% 541068
j,l 0.00% 443193
j,n 0.01% 8902
j,m 0.00% 428334
l,n 0.07% 1436
l,m 0.03% 3620
m,n 0.04% 2737
e,u 0.18% 560
i,y 0.03% 3443
Total 1.13% 88
Original:
digraphs Occurrences Percentage Period
f,e 11259063520 0.3993% 250
y,o 5246910244 0.1861% 537
y,u 166218756 0.0059% 16964
Total 16672192520 0.5913% 169
Post-swap:
digraphs Occurrences Percentage Period
y,e 6666086273 0.2364% 423
f,o 46883347757 1.6627% 60
f,u 3228716759 0.1145% 873
Total 56778150789 2.0137% 50
Update:
Been playing with the "number shifting" layout for a week, and I've found Z and Q are a lot easier to hit than at the center :) Now at around 50-60 wpm, at the point where the text starts to "flow".
Also swapped w/m. The changes are relatively small (especially since it's a same-finger swap), but here goes:
Advantages:
wn, nw digraphs (0.08%) now easier to hit
mb, bm digraphs (0.09%) easier to hit - alternative fingering may now even be used.
Trigraphs containing wg wv md mb (0.60%) are made easier.
Disadvantages:
Increased use of "hard to reach" QWERTY r - even over colemak and dvorak (m vs p and p)
Trigraphs containing mg mv wb wd (0.28%) made worse.
Unknown:
Decreased use of QWERTY c
Putting d at QWERTY v position (say, if one wanted to adhere to the workman model) is made less viable, due to the relatively high number of trigraphs containing dm (0.30%)
tm,mt (0.03%) is slightly less common than tw,wt (0.09%) - whether or not this is an advantage or disadvantage depends on the relative ease of hitting QWERTY f QWERTY r and QWERTY f QWERTY c.
Sidenote: comparing this layout w,g,d,m -> QWERTY c,v,g,r with the "most workman" variant (m,d,g,w):
Left-hand row-jumps (i.e. from bottom to top row or vice versa) are improved to 0.18% [expected every 555 letters], compared to 0.51% [expected every 198 letters]
mb wv mc ml (0.13% digraphs, 1.20% trigraphs) are improved, the main gains coming from mb (0.09% digraphs, 0.26% trigraphs), ml (0.03% digraphs, 0.46% trigraphs), and mc (0.01% digraphs, 0.44% trigraphs)
wd wb gm vm wc wl (0.05% digraphs, 0.66% trigraphs) are worsened, the main losses coming from wl (0.03% digraphs, 0.37% trigraphs), gm (0.01% digraphs, 0.12% trigraphs), wd (0.01% digraphs, 0.10% trigraphs), vm (0.00% digraphs, 0.05% trigraphs)
The d and m positions, in particular, are very un-workman like, and a workman idealist might not be happy with these changes. However, though these numbers may not seem like much, for example, 0.09% mb digraph implies it occurs, on average, once every 769 characters, and 0.26% implies you'll have a trigraph with m and b, on average, every 385 characters. Were m and b on QWERTY c and t respectively, those would still be pretty hard - and roughly agree with my anecdotal annoyance with those trigraphs.
]]>I originally started with the comma-shifting layout, but couldn't find a place to put k that wasn't annoying. So I ended up with the comma-preserving layout.
Added dvorak to the comparisons.
]]>One thing I don't like is the "wide" part. I've shifted the right hand over one with colemak and just don't use the keys left open; I find them too hard to reach. I guess it's not that bad for z and q since they're not used as frequently, but it's pretty uncomfortable to hit the location for k for me at least.
]]>That the layout be "wide", with right-hand home row shifted to the right. Most existing alternating layouts crowd the keys to the right, making a simple displacement difficult.
Roughly satisfying the workman effort model. Most importantly, upward pinkie stretches are considered very hard, upward middle and ring finger stretches are considered relatively easy, and downward index-finger "folds" are considered easier than upward "stretches".
Roughly optimized for my touchtyping technique (as seen in the diagrams)
Notably, good shortcuts were not on the list of priorities.
License: The layouts, all my content in this post, and all of my own content that is linked to under "Implementations" are available under Creative Commons Attribution 4.0 International (CC BY 4.0) (please provide a link to this thread).
This family of layouts contains two main representatives (though other variations of the two are possible; see section below):
Widely Alternating A (comma-preserving) - retains the comma position, has fewer stretches and more curls
Status: Highly experimental, work in progress. This is used as my secondary layout.
Implementations:
xkb template (not ready for public use)
Widely Alternating A (comma-shifting) - reduces curls, but at the cost of stretching to reach the k
Status: Highly experimental, work not in progress. Some changes from comma-preserving layout may be carried over.
Implementations:
xkb template (not ready for public use)
Compared to Colemak:
Advantages:
The "harder" QWERTY u positions is much less frequent (f vs l).
The "harder" QWERTY c,b,t positions are less frequent (w,v,b vs c,b,g).
The "easier" QWERTY v,w,e positions are more frequent (g,c,l vs v,w,f).
Disadvantages:
Fewer rolls than colemak(?), some awkward direction switching
The "harder" QWERTY n,r positions are more frequent (p,m vs k,p)
Unknown:
The "much harder" QWERTY p position is, by default, more frequent ('\ vs ;:)
The home-row QWERTY h position is much less frequent (y vs h)
The "easy stretch" QWERTY i position is much more frequent (o vs u)
Finger travel is 98.1% of Colemak's
Compared to Dvorak:
Advantages:
The "much harder" QWERTY y,p positions are much less frequent (j,' vs f,l)
The "harder" QWERTY t position is less frequent (b vs y).
The "easier" QWERTY v,w,e positions are much more frequent (g,c,l vs k . ,).
92.1% finger travel
Disadvantages:
The "harder" QWERTY n,b,r positions are more frequent (p,v,m vs b,x,p)
The "easier" QWERTY m position is less frequent (k vs m)
Unknown:
The home-row QWERTY h position is much less frequent (y vs d)
The "easy stretch" QWERTY i position is much more frequent (o vs c)
General:
Disadvantages:
Same-hand row-jumping (i.e. top to bottom row or vice versa) is relatively high. For letters only, 1.09% [expected every 92 chars] compared to colemak's 0.50% [expected every 201 chars], or dvorak's 0.87% [expected every 115 chars].
The main culprits are "op" (0.59%, expected every 171 chars) and "up" (0.24%, expected every 416 chars).
Trigraphs containing phy (0.03%)
Subjective opinions: Compared to more typical alternating layouts like HIEAZMTSRN remarkably similar to colemak in typing positions. The increased sideways motion (when compared with workman-emphasizing swaps of m,d,g,w) is noticable.
Variations:
punctuation can be rearranged as needed.
The m,d,g, and to a lesser extent b,v,w positions can be swapped as preferred. In fact, the original version of this layout had g at the QWERTY c position. However, I found the very common ng digram (hit with middle then index finger) uncomfortable in that position.
jJ kK /? -\ ;: and to a lesser extent zz qQ and some of the other punctuation keys can be swapped as preferred.
In the comma-preserving layout, p and k might be swapped. The main reason I put p there is due to the somewhat frequent "po" digraph.
Open questions:
How important is the workman effort model, in comparison to considerations such as digraph/trigraph awkwardness?
More specifically, compare workman-emphasizing schemes such as m,d,g,w -> QWERTY c,v,g,r to the current one. (My comparison with w,g,d,m.). Is there a better workman-emphasizing scheme?
To be continued...
]]>