Timeline for Bug in MovingMap (breaking change between version 10.1 and 10.0.2)?
Current License: CC BY-SA 3.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 16, 2015 at 22:29 | comment | added | Edmund | @StefanR Thank you for clarifying that the omission of the changes from the release notes was accidental. It is also encouraging to hear that the process will be reviewed to reduce the chance of this sort of thing happening in future. Mathematica is a great tool but this is my first upgrade experience while using it professionally. Therefore, my first experience closely tracking version changes as it is no longer a hobby tool for me. This was a scary one. | |
| Apr 16, 2015 at 19:16 | comment | added | Oleksandr R. | Thanks. That seems like a sensible definition, if not consistent with the last one. (I never used MovingMap before, so it didn't affect me anyway.) | |
| Apr 16, 2015 at 18:31 | comment | added | Stefan R | @OleksandrR. Note that MovingMap[f,list,n] will give {f[list[[1;;1+n]]], f[list[[2;;2+n]]], ...}, so for lists the Quantity[_,"Events"] is not required. Also, the "Events" quantity has special meaning in MovingMap. Using "Kilograms" will probably not produce meaningful results, but for TimeSeries data with dates, quantities like Quantity[1, "Weeks"] can be used. | |
| Apr 16, 2015 at 18:25 | comment | added | Stefan R | @Edmund It should definitely have been mentioned in the release notes. We are looking into why it wasn't included, and how we can avoid such omissions in the future. We are sorry for any frustration due to this issue. This was the only major backwards-incompatible redesign in 10.1 in statistics, and I'm not personally aware of other WL functions that were similarly extensively changed. | |
| Apr 16, 2015 at 10:07 | comment | added | Oleksandr R. | Apologies for what may seem like a flippant comment, but what is the value of having a quantity required in the third place? Nowhere else in Mathematica do we refer to raw numbers having units. And what would be the meaning of, for example, MovingMap[..., Quantity[2, "Kilograms"]]? | |
| Apr 15, 2015 at 23:46 | comment | added | Edmund | I thought I might have missed something so double-checked Summary of New Features in 10.1. As I remembered, no mention of changes to MovingMap. This just can't be done! You can't have changes in results for the same parameters on what should be deterministic functions and leave it to pure chance that the developers using your product will find them. It is ludicrous! I'm thinking: What else has changed in this manner? What other production code is broken from something like this? | |
| Apr 15, 2015 at 23:41 | comment | added | Edmund | @StefanR It is a severely poor practice to change the functionality of a feature and make no explicit mention of it to users. MovingMap has been developed into solutions with the existing 10.0.x functionality. WRI deliberately changing it without notice introduces significant logical errors into existing code. Errors that the developer has no reasonable means to identity or narrow as zero information has been provided. I strongly urge WRI to explicitly state what functions have been altered such that legal syntax produces different results. How is anyone to manage without this info? | |
| Apr 15, 2015 at 23:02 | comment | added | Kuba | Any other already documented functions underwent similar procedure?... | |
| Apr 15, 2015 at 20:10 | comment | added | Stefan R | @DavidZhang Yes, there was an oversight here. The "Events" quantity should not be triggering server calls. We're looking into it. | |
| Apr 15, 2015 at 18:28 | comment | added | David Zhang | @IstvánZachar Try Quantity[2, IndependentUnit["events"]] to prevent Mathematica from trying to interpret "events" as a built-in unit. For some reason, "Events" is not a recognized unit in Mathematica 10.1, even though it is used by MovingMap[]. | |
| Apr 15, 2015 at 17:55 | comment | added | Stefan R | @IstvánZachar That is most likely an issue connecting to the WRI servers which perform unit interpretations. | |
| Apr 15, 2015 at 17:12 | comment | added | István Zachar | This returns an error for me (v10.1.0, 64bit, Win7): Quantity::unkunit: Unable to interpret unit specification Events. >> | |
| Apr 15, 2015 at 16:20 | history | answered | Stefan R | CC BY-SA 3.0 |