candela
candela copied to clipboard
Kitware
Eliminate distinction between Number and Integer?
"Integer" is sort of a false distinction in JavaScript, since integers are represented by Numbers. Eliminating the distinction in Resonant Lab would have two benefits:
- The UI would become less cluttered and strictly easier to understand
- The code would become a little bit shorter
For the record, we were using "integer" distinct from "number" because we wanted our types to conform to those in use in another project (I think it was Minerva?), but I can't seem to find the documentation page that listed the types.
Are there any visualizations that treat integers as distinct from numbers or only handle integers?
On Tue, Jun 28, 2016 at 5:05 PM, Alex Bigelow [email protected] wrote:
For the record, we were using "integer" distinct from "number" because we wanted our types to conform to those in use in another project (I think it was Minerva?), but I can't seem to find the documentation page that listed the types.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/Kitware/candela/issues/277#issuecomment-229183331, or mute the thread https://github.com/notifications/unsubscribe/AIX_R9cIk5T5kRhVKJf5volocwp6uY92ks5qQYyggaJpZM4JAhtq .
David Manthey R&D Engineer Kitware Inc. (518) 881-4439
I don't think any of the ones in Candela do, aside from treating an integer column as qualitative (for, e.g., the bar chart to identify distinct bars). But that operation is general with respect to type so doesn't require drawing the Number/Integer distinction.
Yeah, I was thinking the same thing—it's possible a visualization might care about it for qualitative vs quantitative reasons, but if you're trying to use floats as qualitative values, you probably have bigger problems than just type coercion (this is where we'd punt to Resonant Data Reshaper or something).
I think UpSet allows integers, but not numbers, for set identification, but really we're just using integers there as proxies for boolean values (the "correct" way is probably to auto-detect/specify that those values are actually booleans. Type auto-detection currently isn't that smart, but with the histograms available, it might not actually be that hard).
The only other use for it that I can imagine is where you're expecting the data to all be integers, but you actually have a few floats in there. Resonant Lab would probably auto-detect the field as a number instead of an integer, and the histogram would be able to explain why. Without the distinction, you'd have no way of knowing about the floats until you tried to visualize the page of data that contains the weird values.
With regard to clutter / understanding difficulty, I don't see much of a difference—we're already making people differentiate between numbers, booleans, strings, etc... it's just one more category. As all the integer stuff is already implemented (unless there's more work to do that I'm unaware of), my vote would be to leave it in there, and make sure we ask about it in the next round of user testing.
I've already found Resonant Lab useful for detecting bad data in a data set. If I'm expecting integers and the column appears as number, that is quite useful to me, even if it has no other utility.
On Fri, Jul 1, 2016 at 10:43 AM, Alex Bigelow [email protected] wrote:
Yeah, I was thinking the same thing—it's possible a visualization might care about it for qualitative vs quantitative reasons, but if you're trying to use floats as qualitative values, you probably have bigger problems than just type coercion (this is where we'd punt to Resonant Data Reshaper or something).
I think UpSet allows integers, but not numbers, for set identification, but really we're just using integers there as proxies for boolean values (the "correct" way is probably to auto-detect/specify that those values are actually booleans. Type auto-detection currently isn't that smart, but with the histograms available, it might not actually be that hard).
The only other use for it that I can imagine is where you're expecting the data to all be integers, but you actually have a few floats in there. Resonant Lab would probably auto-detect the field as a number instead of an integer, and the histogram would be able to explain why. Without the distinction, you'd have no way of knowing about the floats until you tried to visualize the page of data that contains the weird values.
With regard to clutter / understanding difficulty, I don't see much of a difference—we're already making people differentiate between numbers, booleans, strings, etc... it's just one more category. As all the integer stuff is already implemented (unless there's more work to do that I'm unaware of), my vote would be to leave it in there, and make sure we ask about it in the next round of user testing.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Kitware/candela/issues/277#issuecomment-229965122, or mute the thread https://github.com/notifications/unsubscribe/AIX_R8cUiA7PpoWYu8A0J_GamBckyLHfks5qRSd0gaJpZM4JAhtq .
David Manthey R&D Engineer Kitware Inc. (518) 881-4439
Since we have at least one vote for its usefulness, let's leave it in for now. My concern about clutter isn't the absolute number of icons, it's the visual similarity of the icons combined with (my perceived) lack of "power" in the distinction between the associated types. But let's definitely ask/observe about it in the next round of user testing.
Let's leave the issue open until then, since it feels like that round of user testing will be the final answer on this.
On Fri, Jul 1, 2016 at 10:46 AM David Manthey [email protected] wrote:
I've already found Resonant Lab useful for detecting bad data in a data set. If I'm expecting integers and the column appears as number, that is quite useful to me, even if it has no other utility.
On Fri, Jul 1, 2016 at 10:43 AM, Alex Bigelow [email protected] wrote:
Yeah, I was thinking the same thing—it's possible a visualization might care about it for qualitative vs quantitative reasons, but if you're trying to use floats as qualitative values, you probably have bigger problems than just type coercion (this is where we'd punt to Resonant Data Reshaper or something).
I think UpSet allows integers, but not numbers, for set identification, but really we're just using integers there as proxies for boolean values (the "correct" way is probably to auto-detect/specify that those values are actually booleans. Type auto-detection currently isn't that smart, but with the histograms available, it might not actually be that hard).
The only other use for it that I can imagine is where you're expecting the data to all be integers, but you actually have a few floats in there. Resonant Lab would probably auto-detect the field as a number instead of an integer, and the histogram would be able to explain why. Without the distinction, you'd have no way of knowing about the floats until you tried to visualize the page of data that contains the weird values.
With regard to clutter / understanding difficulty, I don't see much of a difference—we're already making people differentiate between numbers, booleans, strings, etc... it's just one more category. As all the integer stuff is already implemented (unless there's more work to do that I'm unaware of), my vote would be to leave it in there, and make sure we ask about it in the next round of user testing.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Kitware/candela/issues/277#issuecomment-229965122, or mute the thread < https://github.com/notifications/unsubscribe/AIX_R8cUiA7PpoWYu8A0J_GamBckyLHfks5qRSd0gaJpZM4JAhtq
.
David Manthey R&D Engineer Kitware Inc. (518) 881-4439
—
You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Kitware/candela/issues/277#issuecomment-229966138, or mute the thread https://github.com/notifications/unsubscribe/ACxNJEJvk3FkdkRlll5K9y_0ZV8gz7OAks5qRShfgaJpZM4JAhtq .
The \mathbb{R} and \mathbb{Z} glyphs are one idea, though that may be even more obscure than having to learn the current ones.
On Fri, Jul 1, 2016 at 11:02 AM Alex Bigelow [email protected] wrote:
A new icon's always an easy fix—I'm open to suggestions.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Kitware/candela/issues/277#issuecomment-229970003, or mute the thread https://github.com/notifications/unsubscribe/ACxNJHyjumd-PQAUlznsKXr1gr6PAL_eks5qRSvhgaJpZM4JAhtq .