Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[splinterz]

I need help with a filtering script.

I would like to create a filter that lists every Dwarf who has the preference "Likes working outdoors" or "Does not mind being outdoors".

I assume I would create a script using d.has_preference(), but I don't know what to put in the brackets.  How exactly does this function work, and what are the values that I need to use to get my filter working?

what you'll have to use for now is:

Code: [Select]

d.find_preference('Does not mind being outdoors','Outdoors') || d.find_preference('Likes working outdoors','Outdoors')
there seems to be an issue with how the script engine handles (or rather doesn't handle) optional function parameters that i need to look into.

[grody311]

what you'll have to use for now is:

Code: [Select]

d.find_preference('Does not mind being outdoors','Outdoors') || d.find_preference('Likes working outdoors','Outdoors')
there seems to be an issue with how the script engine handles (or rather doesn't handle) optional function parameters that i need to look into.

Perfect.  I don't mind using two separate scripts for now.  Thanks!

So I guess the arguments need to be d.find_preferences('exact preference','preference category').  Good to know.

edit: As a side note, after some experimenting, it appears that using the optional parameter in this case seems to work correctly.

[thistleknot]

I've been lurking, but I've been working on a new algorithm that transforms the %'s to a more "natural" distribution curve.

It's hard to describe... but s transform is fucking awesome IMO.

It just has a "jagged" shape to it's curve.

I am able to measure the change in curves by utilizing trapezoid area rules applied to integral approximation.  I so far am able to achieve some far better results than S Transform.  I was concerned it might look too much like "rank-ecdf" which ignored % gaps between values, but I found my concern was unfounded.  I think it's the best approximation of the [normal] PDF of any given dataset so far.

http://imgur.com/gallery/jyv3kGQ/new

[PeridexisErrant]

If I drew a graph like that without axis labels, the whole report would fail...  (seriously, I have no idea what that is)

[thistleknot]

that's me listing the data probabilities (in a probability density function to each ATTRIBUTE point from 19 observed points) from 0 to 100%

but I subtracted .5

then took the ABSOLUTE VALUE of that

subtracted .5 again!

and graphed it.

seriously, if you want the worksheet, pm me, otherwise I won't bore the readers

[salithus]

that's me listing the data probabilities (in a probability density function to each ATTRIBUTE point from 19 observed points) from 0 to 100%

but I subtracted .5

then took the ABSOLUTE VALUE of that

subtracted .5 again!

and graphed it.

seriously, if you want the worksheet, pm me, otherwise I won't bore the readers

one day I will understand statistics. and then the four riders will come.

[splinterz]

v26.0 patch notes

• 40.11/12/13 memory layouts
• units affected by a werebeast syndrome are properly detected as cursed and will show the type of werebeast
• added belief column type
• added unit kills column type and information in tooltip (optional)
• added a sort method for attribute columns to sort by the maximum potential
• added a right click menu item to update a dwarf's custom profession with their current labors
• gender icons have been updated to also show sexual orientation
• added default noble colors and changed some default settings
• added an option to decorate noble names (italicized with symbol)
• removed general metal/stone/gem/alcohol preferences from all roles
• added new preference groups for shearable, fishable creatures, millable, extract, crop plants
• changed role ratings to display consistently and with a larger spread
• changed the way body sizes are read to hopefully be more accurate
• added retooled default roles (feelotraveller)
• general cleanup, fixes and many back-end optimizations (Hello71)
• fixed an issue with trait/belief conflicts not displaying properly
• fixed custom profession sorting (although it's still case sensitive)
• fixed optimization plans loading/displaying incorrectly in certain cases
• fixed role preference ratings for multiple matches
• fixed active military showing up as 'no job'
• fixed noble color crash (stoutheart)
• fixes for manual and file paths (Hello71)

i haven't had a lot of playtesting with this version, and there's quite a lot of changes, so let me know if anything explodes. i'll also start adding releases on the github page from now on.

[thraxarious]

Awesome! Thanks!

[Aquathug]

Incredible.

Thanks to all the DT devs who keep this project alive and kickin.

[thistleknot]

So turns out all my work on s transform and integral's has lead me full circle.

Spoiler: TLDR (click to show/hide)

Doesn't look like an approximated integral (using trapezoid rule) doesn't yield me any difference in results than from an initial minmax transform.

DT uses a slight variant from a minmax transform, and puts the mean at 50% by doing a minmax transform around the mean for 0 to 50% and 50% to 100%.

Then a similar transform is done again to recenter on the median.

Resulting in [more or less] a nicely distributed set of values.

I was hoping that by approximating "integral's" I would be able to find a better "curve".  It appears that the concept of using the trapezoid rule to approximate integrals produces %'s that are exactly the same as a minmax transform.  I was struggling with why my integral's 50% point did not represent the mean, but rather the midrange: (max - min) /2.

However, I know why [now], because that's what a minmax transform is, and an approximated integral based on the trapezoid rule produces the same #'s.

So I just wasted a whole lot of time going full circle.

However, what I would like to focus on.  Is how to deal with datasets whose overall average doesn't = 50%.

I fear that combining them together will slightly skew the overall weighted average of values.

The data is already normalized <>50% on a 50/50 split. (except for extremely skewed datasets that contain a lot of null values: skills/preferences, but there mean is ~.5)

So I was thinking, since the drawing method has been updated to draw from min/max of weighted average outputs, and 50% = median.

Why not do something similar, but instead of adding up the %'s as if they are 0 to 100%.

Subtract .5 from each %
and get a set of values that are
~-50% to ~+50%
but has a perfect even average of 0.

update
you know what, I don't think that would do anything.  One would have to subtract from the datasets mean, and that would have a different issue

Since we end up doing a final transform on the outputted data, and it seems that the #'s on the backend never really add up to 0 or 100% (hence why the new drawing method is used).  We can still work with these #'s, albeit in a slightly better mean adjusted approach.

[ResMar]

Spint, any chance you can do a brief writeup of role calculation methodology? I'm seeing a lot of interest in that as late, and it's something I would like the understand better for future edits to the documentation, but I'm probably not going to able to go off of the conversational back-and-forth going on here.

[khearn]

Thanks for the new release. Looks like lots of interesting new stuff. Thanks especially for the body size change. I'll take a look and see if it comes up with reasonable looking numbers.

   Keith

[lethosor]

i haven't had a lot of playtesting with this version, and there's quite a lot of changes, so let me know if anything explodes. i'll also start adding releases on the github page from now on.

It's difficult to tell what "Dwarf-Therapist.26.0.0.zip" is from the filename (I'm assuming it's a compiled binary, in which case specifying the platform would be helpful).

[Hello71]

[now how do I delete this...]

[khearn]

i haven't had a lot of playtesting with this version, and there's quite a lot of changes, so let me know if anything explodes. i'll also start adding releases on the github page from now on.

It's difficult to tell what "Dwarf-Therapist.26.0.0.zip" is from the filename (I'm assuming it's a compiled binary, in which case specifying the platform would be helpful).

Agreed. I looked in it and it contains a .exe (along with other files), so I assume it's just a windows release. I had no problems compiling for linux from the source .zip file.

Now if I could only figure out how to get git to give me the correct dfhack sources so I could compile it. But that's not germane to this thread.

   Keith