# Follow-up on subsetting data.table with NAs

22 messages
12
Open this post in threaded view
|
Report Content as Inappropriate

## Follow-up on subsetting data.table with NAs

 Matthew, Regarding your recent answer here: http://stackoverflow.com/a/17008872/559784 I'd a few questions/thoughts and I thought it may be more appropriate to share here (even though I've already written 3 comments!).1) First, you write that, DT[ColA == ColB] is simpler than DF[!is.na(ColA) & !is.na(ColB) & ColA == ColB,]However, you can write this long expression as: DF[which(DF\$ColA == DF\$ColB), ]2) Second, you mention that the motivation is not just convenience but speed. By checking:require(data.table)set.seed(45)df <- as.data.frame(matrix(sample(c(1,2,3,NA), 2e6, replace=TRUE), ncol=2))dt <- data.table(df)system.time(dt[V1 == V2])# 0.077 secondssystem.time(df[!is.na(df\$V1) & !is.na(df\$V2) & df\$V1 == df\$V2, ])# 0.252 secondssystem.time(df[which(df\$V1 == df\$V2), ])# 0.038 secondsWe see that using `which` (in addition to removing NA) is also faster than `DT[V1 == V2]`. In fact, `DT[which(V1 == V2)]` is faster than `DT[V1 == V2]`. I suspect this is because of the snippet below in `[.data.table`:        if (is.logical(i)) {            if (identical(i,NA)) i = NA_integer_  # see DT[NA] thread re recycling of NA logical            else i[is.na(i)] = FALSE              # avoids DT[!is.na(ColA) & !is.na(ColB) & ColA==ColB], just DT[ColA==ColB]        }But at the end `irows <- which(i)` is being done:            if (is.logical(i)) {                if (length(i)==nrow(x)) irows=which(i)   # e.g. DT[colA>3,which=TRUE]And this "irows" is what's used to index the corresponding rows. So, is the replacement of `NA` to FALSE really necessary? I may very well have overlooked the purpose of the NA replacement to FALSE for other scenarios, but just by looking at this case, it doesn't seem like it's necessary as you fetch index/row numbers later.3) And finally, more of a philosophical point. If we agree that subsetting can be done conveniently (using "which") and with no loss of speed (again using "which"), then are there other reasons to change the default behaviour of R's philosophy of handling NAs as unknowns/missing observations? I find I can relate more to the native concept of handling NAs. For example:x <- c(1,2,3,NA)x != 3# TRUE TRUE FALSE NAmakes more sense because `NA != 3` doesn't fall in either TRUE or FALSE, if NA is a missing observation/unknown data. The answer "unknown/missing" seems more appropriate, therefore.I'd be interested in hearing, in addition to Matthew's, other's thoughts and inputs as well.Best regards,Arun _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 On 09.06.2013 22:08, Arunkumar Srinivasan wrote: Matthew, Regarding your recent answer here: http://stackoverflow.com/a/17008872/559784 I'd a few questions/thoughts and I thought it may be more appropriate to share here (even though I've already written 3 comments!). 1) First, you write that, DT[ColA == ColB] is simpler than DF[!is.na(ColA) & !is.na(ColB) & ColA == ColB,] However, you can write this long expression as: DF[which(DF\$ColA == DF\$ColB), ] Good point. But DT[ColA == ColB] still seems simpler than DF[which(DF\$ColA == DF\$ColB), ]  (in data.table  DT[which(ColA == ColB)]).   I worry about forgetting I need which() and then have bugs occur when NA occur in the data at some time in future that don't occur now or in test. 2) Second, you mention that the motivation is not just convenience but speed. By checking: require(data.table) set.seed(45) df <- as.data.frame(matrix(sample(c(1,2,3,NA), 2e6, replace=TRUE), ncol=2)) dt <- data.table(df) system.time(dt[V1 == V2]) # 0.077 seconds system.time(df[!is.na(df\$V1) & !is.na(df\$V2) & df\$V1 == df\$V2, ]) # 0.252 seconds system.time(df[which(df\$V1 == df\$V2), ]) # 0.038 seconds We see that using `which` (in addition to removing NA) is also faster than `DT[V1 == V2]`. In fact, `DT[which(V1 == V2)]` is faster than `DT[V1 == V2]`. I suspect this is because of the snippet below in `[.data.table`:         if (is.logical(i)) {             if (identical(i,NA)) i = NA_integer_  # see DT[NA] thread re recycling of NA logical             else i[is.na(i)] = FALSE              # avoids DT[!is.na(ColA) & !is.na(ColB) & ColA==ColB], just DT[ColA==ColB]         } But at the end `irows <- which(i)` is being done:             if (is.logical(i)) {                 if (length(i)==nrow(x)) irows=which(i)   # e.g. DT[colA>3,which=TRUE] And this "irows" is what's used to index the corresponding rows. So, is the replacement of `NA` to FALSE really necessary? I may very well have overlooked the purpose of the NA replacement to FALSE for other scenarios, but just by looking at this case, it doesn't seem like it's necessary as you fetch index/row numbers later. Interesting.  Cool, so dt[V1 == V2] can and should be at least as fast as the which() way.  Will file a FR to improve that speed! 3) And finally, more of a philosophical point. If we agree that subsetting can be done conveniently (using "which") and with no loss of speed (again using "which"), Not sure that is agreed yet, but happy to be persuaded. then are there other reasons to change the default behaviour of R's philosophy of handling NAs as unknowns/missing observations? I find I can relate more to the native concept of handling NAs. For example: x <- c(1,2,3,NA) x != 3 # TRUE TRUE FALSE NA makes more sense because `NA != 3` doesn't fall in either TRUE or FALSE, if NA is a missing observation/unknown data. The answer "unknown/missing" seems more appropriate, therefore. True but the context of where that result is used is all important; i.e., in this case that's `i` of [.data.table or [.data.frame.  It may be easier to consider == first.  The data.table philosophy is that DT [ x==3 ]  should exclude any rows in x that are NA,  without needing to do anything special such as needing to know to call which() as well.  That differs to data.frame,  but is more consistent with SQL.  In SQL "where x = 3" doesn't need anything else if x contains some NULL values. I'd be interested in hearing, in addition to Matthew's, other's thoughts and inputs as well. Best regards, Arun     _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 Matthew,Regarding your suggestion of changes regarding Frank's post here: http://stackoverflow.com/a/17008872/559784 I find it a bit more confusing and frankly not like sql.You wrote: "If I haven't understood correctly feel free to correct, otherwise the change will get made eventually. It will need to be done in a way that considers compound expressions; e.g., `DT[colA=="foo" & colB!="bar"]` should exclude rows with `NA` in `colA` but include rows where `colA` is non-`NA` but `colB` is `NA`. Similarly, `DT[colA!=colB]` should include rows where either colA or colB is `NA` but not both. And perhaps `DT[colA==colB]` should include rows where both`colA` and `colB` are `NA` (which it doesn't currently, I believe)."Even though sql (ex: sqldf) has a different way of handling NAs when compared to data.frame, it doesn't seem to find NA == NA. That is,df <- data.frame(x = c(1:3,NA), y = c(NA,4:5,NA))require(sqldf)sqldf("select * from df where x == y")# returns empty data.framesqldf("select * from df where x != y")  x y1 2 42 3 5That is, at least in sqldf package, NA is not == NA and NA is not != NA which is very much in coherence with R's default NA == NA and NA != NA (both giving NA). But I don't think they it's considered FALSE here. It just acts like the "subset" function where all entries that were evaluated to NAs are simply dropped. But with data.table philosophy NA != NA should be evaluated to TRUE, which I don't think (from what I meagrely understand from sql) is what sql does. Please correct me if I've got it wrong.I think it is clearer and simpler if "NAs are just dropped" after evaluating logical expressions. It would be also easy to document this and easier to grasp, imho. This would also explain Frank's post for NA rows being removed. And probably if there is more consensus an option for "na.rm = TRUE/FALSE" could be added?Arun _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 Hi Arun, Hm, good point.  Is data.table consistent with SQL already, for both == and !=, and so no change needed?  And it was correct for Frank to be mistaken.  Maybe just some more documentation and examples needed then. Are you happy that DT[!(x==.)] and DT[x!=.] do treat NA inconsistently? : http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently "na.rm = TRUE/FALSE" sounds good to me.  I'd only considered nomatch before in the context of joins, not logical subsets.   Thanks, Matthew   On 10.06.2013 08:11, Arunkumar Srinivasan wrote: Matthew, Regarding your suggestion of changes regarding Frank's post here: http://stackoverflow.com/a/17008872/559784 I find it a bit more confusing and frankly not like sql. You wrote: "If I haven't understood correctly feel free to correct, otherwise the change will get made eventually. It will need to be done in a way that considers compound expressions; e.g., `DT[colA=="foo" & colB!="bar"]` should exclude rows with `NA` in `colA` but include rows where `colA` is non-`NA` but `colB` is `NA`. Similarly, `DT[colA!=colB]` should include rows where either colA or colB is `NA` but not both. And perhaps `DT[colA==colB]` should include rows where both`colA` and `colB` are `NA` (which it doesn't currently, I believe)." Even though sql (ex: sqldf) has a different way of handling NAs when compared to data.frame, it doesn't seem to find NA == NA. That is, df <- data.frame(x = c(1:3,NA), y = c(NA,4:5,NA)) require(sqldf) sqldf("select * from df where x == y") # returns empty data.frame sqldf("select * from df where x != y")   x y 1 2 4 2 3 5 That is, at least in sqldf package, NA is not == NA and NA is not != NA which is very much in coherence with R's default NA == NA and NA != NA (both giving NA). But I don't think they it's considered FALSE here. It just acts like the "subset" function where all entries that were evaluated to NAs are simply dropped. But with data.table philosophy NA != NA should be evaluated to TRUE, which I don't think (from what I meagrely understand from sql) is what sql does. Please correct me if I've got it wrong. I think it is clearer and simpler if "NAs are just dropped" after evaluating logical expressions. It would be also easy to document this and easier to grasp, imho. This would also explain Frank's post for NA rows being removed.  And probably if there is more consensus an option for "na.rm = TRUE/FALSE" could be added? Arun     _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 Hm, good point.  Is data.table consistent with SQL already, for both == and !=, and so no change needed?  Yes, I believe it's already consistent with SQL. However, the current interpretation of NA (documentation) being treated as FALSE is not needed / untrue, imho (Please see below). And it was correct for Frank to be mistaken.  Yes, it seems like he was mistaken.Maybe just some more documentation and examples needed then.It'd be much more appropriate if the documentation reflects the role of subsetting in data.table mimicking "subset" function (in order to be in line with SQL) by dropping NA evaluated logicals. From a couple of posts before, where I pasted the code where NAs are replaced to FALSE were not necessary as `irows <- which(i)` makes clear that `which` is being used to get indices and then subset, this fits perfectly well with the interpretation of NA in data.table. Are you happy that DT[!(x==.)] and DT[x!=.] do treat NA inconsistently? : http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently Ha, I like the idea behind the use of () in evaluating expressions. It's another nice layer towards simplicity in data.table. But I still think there should not be an inconsistency in equivalent logical operations to provide different results. If !(x== .) and x != . are indeed different, then I'd suppose replacing `!` with a more appropriate name as it's much easier to get confused otherwise. In essence, either !(x == .) must evaluate to (x != .) if the underlying meaning of these are the same, or the `!` in `!(x==.)` must be replaced to something that's more appropriate for what it's supposed to be. Personally, I prefer the former. It would greatly tighten the structure and consistency. "na.rm = TRUE/FALSE" sounds good to me.  I'd only considered nomatch before in the context of joins, not logical subsets.Yes, I find this option would give more control in evaluating expressions with ease in `i`, by providing both "subset" (default) and the typical data.frame subsetting (na.rm = FALSE).Best regards, Arun _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 Hi Matthew,My view (from the last reply) more or less reflects mnel's comments here: http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently#comment23317096_16240143  Pasted here for convenience:`data.table` is mimicing `subset` in its handling of `NA` values in logical `i` arguments. -- the only issue is the `!` prefix signifying a not-join, not the way one might expect. Perhaps the not join prefix could have been `NJ` not `!` to avoid this confusion -- this might be another discussion to have on the mailing list -- (I think it is a discussion worth having) Arun On Monday, June 10, 2013 at 10:28 AM, Arunkumar Srinivasan wrote: Hm, good point.  Is data.table consistent with SQL already, for both == and !=, and so no change needed?  Yes, I believe it's already consistent with SQL. However, the current interpretation of NA (documentation) being treated as FALSE is not needed / untrue, imho (Please see below). And it was correct for Frank to be mistaken.  Yes, it seems like he was mistaken.Maybe just some more documentation and examples needed then.It'd be much more appropriate if the documentation reflects the role of subsetting in data.table mimicking "subset" function (in order to be in line with SQL) by dropping NA evaluated logicals. From a couple of posts before, where I pasted the code where NAs are replaced to FALSE were not necessary as `irows <- which(i)` makes clear that `which` is being used to get indices and then subset, this fits perfectly well with the interpretation of NA in data.table. Are you happy that DT[!(x==.)] and DT[x!=.] do treat NA inconsistently? : http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently Ha, I like the idea behind the use of () in evaluating expressions. It's another nice layer towards simplicity in data.table. But I still think there should not be an inconsistency in equivalent logical operations to provide different results. If !(x== .) and x != . are indeed different, then I'd suppose replacing `!` with a more appropriate name as it's much easier to get confused otherwise. In essence, either !(x == .) must evaluate to (x != .) if the underlying meaning of these are the same, or the `!` in `!(x==.)` must be replaced to something that's more appropriate for what it's supposed to be. Personally, I prefer the former. It would greatly tighten the structure and consistency. "na.rm = TRUE/FALSE" sounds good to me.  I'd only considered nomatch before in the context of joins, not logical subsets.Yes, I find this option would give more control in evaluating expressions with ease in `i`, by providing both "subset" (default) and the typical data.frame subsetting (na.rm = FALSE).Best regards, Arun _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 Hi, How about ~ instead of ! ?      I ruled out - previously to leave + and - available for future use.  NJ() may be possible too. Matthew   On 10.06.2013 09:35, Arunkumar Srinivasan wrote: Hi Matthew, My view (from the last reply) more or less reflects mnel's comments here: http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently#comment23317096_16240143  Pasted here for convenience: `data.table` is mimicing `subset` in its handling of `NA` values in logical `i` arguments. -- the only issue is the `!` prefix signifying a not-join, not the way one might expect. Perhaps the not join prefix could have been `NJ` not `!` to avoid this confusion -- this might be another discussion to have on the mailing list -- (I think it is a discussion worth having)  Arun On Monday, June 10, 2013 at 10:28 AM, Arunkumar Srinivasan wrote: Hm, good point.  Is data.table consistent with SQL already, for both == and !=, and so no change needed?   Yes, I believe it's already consistent with SQL. However, the current interpretation of NA (documentation) being treated as FALSE is not needed / untrue, imho (Please see below).   And it was correct for Frank to be mistaken.   Yes, it seems like he was mistaken. Maybe just some more documentation and examples needed then. It'd be much more appropriate if the documentation reflects the role of subsetting in data.table mimicking "subset" function (in order to be in line with SQL) by dropping NA evaluated logicals. From a couple of posts before, where I pasted the code where NAs are replaced to FALSE were not necessary as `irows <- which(i)` makes clear that `which` is being used to get indices and then subset, this fits perfectly well with the interpretation of NA in data.table.  Are you happy that DT[!(x==.)] and DT[x!=.] do treat NA inconsistently? : http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently  Ha, I like the idea behind the use of () in evaluating expressions. It's another nice layer towards simplicity in data.table. But I still think there should not be an inconsistency in equivalent logical operations to provide different results. If !(x== .) and x != . are indeed different, then I'd suppose replacing `!` with a more appropriate name as it's much easier to get confused otherwise.  In essence, either !(x == .) must evaluate to (x != .) if the underlying meaning of these are the same, or the `!` in `!(x==.)` must be replaced to something that's more appropriate for what it's supposed to be. Personally, I prefer the former. It would greatly tighten the structure and consistency. "na.rm = TRUE/FALSE" sounds good to me.  I'd only considered nomatch before in the context of joins, not logical subsets. Yes, I find this option would give more control in evaluating expressions with ease in `i`, by providing both "subset" (default) and the typical data.frame subsetting (na.rm = FALSE). Best regards,   Arun     _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 In reply to this post by Arunkumar Srinivasan (Sorry @Matthew for the double email, I forgot to include the list once again).However, one inconsistency I find with the use of `!(x==.)` is this:dt1 <- data.table(x = 0:4, y=5:9)> dt1[!(x)]   x  y1: 4 9Not the correct result! If `!(x==.)` is equal to `x != .`, then the correct result should be the first row, isn't it?dt2 <- data.table(x = c(0,3,4,NA), y = c(NA,4,5,NA))> dt2[!(x)] # ends up in an errorError in seq_len(nrow(x))[-irows] :   only 0's may be mixed with negative subscriptsIt ends up in an error because `NA` is not removed/replaced.Running the same on data.frame gives the results it's supposed to.Arun _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 In reply to this post by Arunkumar Srinivasan   On 10.06.2013 09:53, Arunkumar Srinivasan wrote: However, one inconsistency I find with the use of `!(x==.)` is this: dt1 <- data.table(x = 0:4, y=5:9) > dt1[!(x)]    x  y 1: 4 10 Not the correct result! If `!(x==.)` is equal to `x != .`, then the correct result should be the first row, isn't it? That result makes perfect sense to me.   I don't think of !(x==.) being the same as  x!=.    ! is simply a prefix.    It's all the rows that aren't returned if the ! prefix wasn't there. dt2 <- data.table(x = c(0,3,4,NA), y = c(NA,4,5,NA)) > dt2[!(x)] # ends up in an error Error in seq_len(nrow(x))[-irows] :    only 0's may be mixed with negative subscripts That needs to be fixed.  But we're getting quite theoretical here and far away from common use cases.  Why would we ever have row numbers of the table, as a column of the table itself and want to select the rows by number not mentioned in that column? It ends up in an error because `NA` is not removed/replaced. Running the same on data.frame gives the results it's supposed to. Arun On Monday, June 10, 2013 at 10:35 AM, Arunkumar Srinivasan wrote: Hi Matthew, My view (from the last reply) more or less reflects mnel's comments here: http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently#comment23317096_16240143 Pasted here for convenience: `data.table` is mimicing `subset` in its handling of `NA` values in logical `i` arguments. -- the only issue is the `!` prefix signifying a not-join, not the way one might expect. Perhaps the not join prefix could have been `NJ` not `!` to avoid this confusion -- this might be another discussion to have on the mailing list -- (I think it is a discussion worth having)  Arun On Monday, June 10, 2013 at 10:28 AM, Arunkumar Srinivasan wrote: Hm, good point.  Is data.table consistent with SQL already, for both == and !=, and so no change needed?   Yes, I believe it's already consistent with SQL. However, the current interpretation of NA (documentation) being treated as FALSE is not needed / untrue, imho (Please see below). And it was correct for Frank to be mistaken.   Yes, it seems like he was mistaken. Maybe just some more documentation and examples needed then. It'd be much more appropriate if the documentation reflects the role of subsetting in data.table mimicking "subset" function (in order to be in line with SQL) by dropping NA evaluated logicals. From a couple of posts before, where I pasted the code where NAs are replaced to FALSE were not necessary as `irows <- which(i)` makes clear that `which` is being used to get indices and then subset, this fits perfectly well with the interpretation of NA in data.table.  Are you happy that DT[!(x==.)] and DT[x!=.] do treat NA inconsistently? : http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently  Ha, I like the idea behind the use of () in evaluating expressions. It's another nice layer towards simplicity in data.table. But I still think there should not be an inconsistency in equivalent logical operations to provide different results. If !(x== .) and x != . are indeed different, then I'd suppose replacing `!` with a more appropriate name as it's much easier to get confused otherwise.  In essence, either !(x == .) must evaluate to (x != .) if the underlying meaning of these are the same, or the `!` in `!(x==.)` must be replaced to something that's more appropriate for what it's supposed to be. Personally, I prefer the former. It would greatly tighten the structure and consistency. "na.rm = TRUE/FALSE" sounds good to me.  I'd only considered nomatch before in the context of joins, not logical subsets. Yes, I find this option would give more control in evaluating expressions with ease in `i`, by providing both "subset" (default) and the typical data.frame subsetting (na.rm = FALSE). Best regards, Arun     _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 In reply to this post by Matthew Dowle Matthew, How about ~ instead of ! ?      I ruled out - previously to leave + and - available for future use.  NJ() may be possible too.Both "NJ()" and "~" are okay for me.That result makes perfect sense to me.   I don't think of !(x==.) being the same as  x!=.    ! is simply a prefix.    It's all the rows that aren't returned if the ! prefix wasn't there.I understand that `DT[!(x)]` does what `data.table` is designed to do currently. What I failed to mention was that if one were to consider implementing `!(x==.)` as the same as `x != .` then this behaviour has to be changed. Let's forget this point for a moment.That needs to be fixed.  But we're getting quite theoretical here and far away from common use cases.  Why would we ever have row numbers of the table, as a column of the table itself and want to select the rows by number not mentioned in that column?Probably I did not choose a good example. Suppose that I've a data.table and I want to get all rows where "x == 0". Let's say:set.seed(45)DT <- data.table( x = sample(c(0,5,10,15), 10, replace=TRUE), y = sample(15)) DF <- as.data.frame(DT)To get all rows where x == 0, it could be done with DT[x == 0]. But it makes sense, at least in the context of data.frames, to do equivalently,DF[!(DF\$x), ] (or) DF[DF\$x == 0, ]All I want to say is, I expect `DT[!(x)]` should give the same result as `DT[x == 0]` (even though I fully understand it's not the intended behaviour of data.table), as it's more intuitive and less confusing. So, changing `!` to `~` or `NJ` is one half of the issue for me. The other is to replace the actual function of `!` in all contexts. I hope I came across with what I wanted to say, better this time.Best,Arun On Monday, June 10, 2013 at 10:52 AM, Matthew Dowle wrote:   Hi, How about ~ instead of ! ?      I ruled out - previously to leave + and - available for future use.  NJ() may be possible too. Matthew   On 10.06.2013 09:35, Arunkumar Srinivasan wrote: Hi Matthew, My view (from the last reply) more or less reflects mnel's comments here: http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently#comment23317096_16240143  Pasted here for convenience: `data.table` is mimicing `subset` in its handling of `NA` values in logical `i` arguments. -- the only issue is the `!` prefix signifying a not-join, not the way one might expect. Perhaps the not join prefix could have been `NJ` not `!` to avoid this confusion -- this might be another discussion to have on the mailing list -- (I think it is a discussion worth having)  Arun On Monday, June 10, 2013 at 10:28 AM, Arunkumar Srinivasan wrote: Hm, good point.  Is data.table consistent with SQL already, for both == and !=, and so no change needed?   Yes, I believe it's already consistent with SQL. However, the current interpretation of NA (documentation) being treated as FALSE is not needed / untrue, imho (Please see below).   And it was correct for Frank to be mistaken.   Yes, it seems like he was mistaken. Maybe just some more documentation and examples needed then. It'd be much more appropriate if the documentation reflects the role of subsetting in data.table mimicking "subset" function (in order to be in line with SQL) by dropping NA evaluated logicals. From a couple of posts before, where I pasted the code where NAs are replaced to FALSE were not necessary as `irows <- which(i)` makes clear that `which` is being used to get indices and then subset, this fits perfectly well with the interpretation of NA in data.table.  Are you happy that DT[!(x==.)] and DT[x!=.] do treat NA inconsistently? : http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently  Ha, I like the idea behind the use of () in evaluating expressions. It's another nice layer towards simplicity in data.table. But I still think there should not be an inconsistency in equivalent logical operations to provide different results. If !(x== .) and x != . are indeed different, then I'd suppose replacing `!` with a more appropriate name as it's much easier to get confused otherwise.  In essence, either !(x == .) must evaluate to (x != .) if the underlying meaning of these are the same, or the `!` in `!(x==.)` must be replaced to something that's more appropriate for what it's supposed to be. Personally, I prefer the former. It would greatly tighten the structure and consistency. "na.rm = TRUE/FALSE" sounds good to me.  I'd only considered nomatch before in the context of joins, not logical subsets. Yes, I find this option would give more control in evaluating expressions with ease in `i`, by providing both "subset" (default) and the typical data.frame subsetting (na.rm = FALSE). Best regards,   Arun     _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 +1 to using ~ for the not-join/join on complement/complement then join. Having some logical-looking i's lead to subsetting and others to not-joins can (for me) lead to mistakes that I'm not likely to catch until much later, if at all. I'm not sure I follow Arun's second example. If the syntax is changed so that ~ works as ! does now, then presumably !x will be reverted to having only a logical interpretation -- coercing x to logical and taking the subset where x == 0 -- which is the behavior you want. So why is it a separate issue? The remaining difference from data.frames would be that DF[!x] would show NA rows, if any, while DT[!x] would not. --FrankOn Mon, Jun 10, 2013 at 4:21 AM, Arunkumar Srinivasan wrote: Matthew, How about ~ instead of ! ?      I ruled out - previously to leave + and - available for future use.  NJ() may be possible too. Both "NJ()" and "~" are okay for me.That result makes perfect sense to me.   I don't think of !(x==.) being the same as  x!=.    ! is simply a prefix.    It's all the rows that aren't returned if the ! prefix wasn't there. I understand that `DT[!(x)]` does what `data.table` is designed to do currently. What I failed to mention was that if one were to consider implementing `!(x==.)` as the same as `x != .` then this behaviour has to be changed. Let's forget this point for a moment. That needs to be fixed.  But we're getting quite theoretical here and far away from common use cases.  Why would we ever have row numbers of the table, as a column of the table itself and want to select the rows by number not mentioned in that column? Probably I did not choose a good example. Suppose that I've a data.table and I want to get all rows where "x == 0". Let's say:set.seed(45)DT <- data.table( x = sample(c(0,5,10,15), 10, replace=TRUE), y = sample(15))  DF <- as.data.frame(DT)To get all rows where x == 0, it could be done with DT[x == 0]. But it makes sense, at least in the context of data.frames, to do equivalently, DF[!(DF\$x), ] (or) DF[DF\$x == 0, ]All I want to say is, I expect `DT[!(x)]` should give the same result as `DT[x == 0]` (even though I fully understand it's not the intended behaviour of data.table), as it's more intuitive and less confusing.  So, changing `!` to `~` or `NJ` is one half of the issue for me. The other is to replace the actual function of `!` in all contexts. I hope I came across with what I wanted to say, better this time. Best,Arun On Monday, June 10, 2013 at 10:52 AM, Matthew Dowle wrote:   Hi, How about ~ instead of ! ?      I ruled out - previously to leave + and - available for future use.  NJ() may be possible too. Matthew   On 10.06.2013 09:35, Arunkumar Srinivasan wrote: Hi Matthew, My view (from the last reply) more or less reflects mnel's comments here: http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently#comment23317096_16240143  Pasted here for convenience: `data.table` is mimicing `subset` in its handling of `NA` values in logical `i` arguments. -- the only issue is the `!` prefix signifying a not-join, not the way one might expect. Perhaps the not join prefix could have been `NJ` not `!` to avoid this confusion -- this might be another discussion to have on the mailing list -- (I think it is a discussion worth having)  Arun On Monday, June 10, 2013 at 10:28 AM, Arunkumar Srinivasan wrote: Hm, good point.  Is data.table consistent with SQL already, for both == and !=, and so no change needed?   Yes, I believe it's already consistent with SQL. However, the current interpretation of NA (documentation) being treated as FALSE is not needed / untrue, imho (Please see below).   And it was correct for Frank to be mistaken.   Yes, it seems like he was mistaken. Maybe just some more documentation and examples needed then. It'd be much more appropriate if the documentation reflects the role of subsetting in data.table mimicking "subset" function (in order to be in line with SQL) by dropping NA evaluated logicals. From a couple of posts before, where I pasted the code where NAs are replaced to FALSE were not necessary as `irows <- which(i)` makes clear that `which` is being used to get indices and then subset, this fits perfectly well with the interpretation of NA in data.table.  Are you happy that DT[!(x==.)] and DT[x!=.] do treat NA inconsistently? : http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently  Ha, I like the idea behind the use of () in evaluating expressions. It's another nice layer towards simplicity in data.table. But I still think there should not be an inconsistency in equivalent logical operations to provide different results. If !(x== .) and x != . are indeed different, then I'd suppose replacing `!` with a more appropriate name as it's much easier to get confused otherwise.  In essence, either !(x == .) must evaluate to (x != .) if the underlying meaning of these are the same, or the `!` in `!(x==.)` must be replaced to something that's more appropriate for what it's supposed to be. Personally, I prefer the former. It would greatly tighten the structure and consistency. "na.rm = TRUE/FALSE" sounds good to me.  I'd only considered nomatch before in the context of joins, not logical subsets. Yes, I find this option would give more control in evaluating expressions with ease in `i`, by providing both "subset" (default) and the typical data.frame subsetting (na.rm = FALSE). Best regards,   Arun     _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

 Frank, You're right about my final point. I can't recollect why I wrote that now. I guess the `!` function will be restored automatically.With my second example, all I wanted to establish was that there was another reason to change `!` from performing the action of a "Not Join" because `DT[!x]` is a perfectly valid syntax (for those who have worked with data.frames and have shifted to data.table) which will not perform the intended action as it'll be a Not Join. In addition, `DT[!x]` gives an error when "x" column has NA. This was meant to be an additional argument for not having `!` for Not Join. But this has caused more confusion. Let's forget about my examples :).To conclude, "~" or "NJ" makes sense than `!` for "Not join" and of course the function of `!` will be automatically restored to "not" (also preferably with a na.rm = TRUE/FALSE. This is what I intended to say from the original discussion. Sorry for any confusion. Arun On Monday, June 10, 2013 at 3:20 PM, Frank Erickson wrote: +1 to using ~ for the not-join/join on complement/complement then join. Having some logical-looking i's lead to subsetting and others to not-joins can (for me) lead to mistakes that I'm not likely to catch until much later, if at all. I'm not sure I follow Arun's second example. If the syntax is changed so that ~ works as ! does now, then presumably !x will be reverted to having only a logical interpretation -- coercing x to logical and taking the subset where x == 0 -- which is the behavior you want. So why is it a separate issue? The remaining difference from data.frames would be that DF[!x] would show NA rows, if any, while DT[!x] would not. --FrankOn Mon, Jun 10, 2013 at 4:21 AM, Arunkumar Srinivasan wrote: Matthew, How about ~ instead of ! ?      I ruled out - previously to leave + and - available for future use.  NJ() may be possible too. Both "NJ()" and "~" are okay for me.That result makes perfect sense to me.   I don't think of !(x==.) being the same as  x!=.    ! is simply a prefix.    It's all the rows that aren't returned if the ! prefix wasn't there.I understand that `DT[!(x)]` does what `data.table` is designed to do currently. What I failed to mention was that if one were to consider implementing `!(x==.)` as the same as `x != .` then this behaviour has to be changed. Let's forget this point for a moment. That needs to be fixed.  But we're getting quite theoretical here and far away from common use cases.  Why would we ever have row numbers of the table, as a column of the table itself and want to select the rows by number not mentioned in that column? Probably I did not choose a good example. Suppose that I've a data.table and I want to get all rows where "x == 0". Let's say:set.seed(45)DT <- data.table( x = sample(c(0,5,10,15), 10, replace=TRUE), y = sample(15))  DF <- as.data.frame(DT)To get all rows where x == 0, it could be done with DT[x == 0]. But it makes sense, at least in the context of data.frames, to do equivalently, DF[!(DF\$x), ] (or) DF[DF\$x == 0, ]All I want to say is, I expect `DT[!(x)]` should give the same result as `DT[x == 0]` (even though I fully understand it's not the intended behaviour of data.table), as it's more intuitive and less confusing.  So, changing `!` to `~` or `NJ` is one half of the issue for me. The other is to replace the actual function of `!` in all contexts. I hope I came across with what I wanted to say, better this time. Best,Arun On Monday, June 10, 2013 at 10:52 AM, Matthew Dowle wrote:   Hi, How about ~ instead of ! ?      I ruled out - previously to leave + and - available for future use.  NJ() may be possible too. Matthew   On 10.06.2013 09:35, Arunkumar Srinivasan wrote: Hi Matthew, My view (from the last reply) more or less reflects mnel's comments here: http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently#comment23317096_16240143  Pasted here for convenience: `data.table` is mimicing `subset` in its handling of `NA` values in logical `i` arguments. -- the only issue is the `!` prefix signifying a not-join, not the way one might expect. Perhaps the not join prefix could have been `NJ` not `!` to avoid this confusion -- this might be another discussion to have on the mailing list -- (I think it is a discussion worth having)  Arun On Monday, June 10, 2013 at 10:28 AM, Arunkumar Srinivasan wrote: Hm, good point.  Is data.table consistent with SQL already, for both == and !=, and so no change needed?   Yes, I believe it's already consistent with SQL. However, the current interpretation of NA (documentation) being treated as FALSE is not needed / untrue, imho (Please see below).   And it was correct for Frank to be mistaken.   Yes, it seems like he was mistaken. Maybe just some more documentation and examples needed then. It'd be much more appropriate if the documentation reflects the role of subsetting in data.table mimicking "subset" function (in order to be in line with SQL) by dropping NA evaluated logicals. From a couple of posts before, where I pasted the code where NAs are replaced to FALSE were not necessary as `irows <- which(i)` makes clear that `which` is being used to get indices and then subset, this fits perfectly well with the interpretation of NA in data.table.  Are you happy that DT[!(x==.)] and DT[x!=.] do treat NA inconsistently? : http://stackoverflow.com/questions/16239153/dtx-and-dtx-treat-na-in-x-inconsistently  Ha, I like the idea behind the use of () in evaluating expressions. It's another nice layer towards simplicity in data.table. But I still think there should not be an inconsistency in equivalent logical operations to provide different results. If !(x== .) and x != . are indeed different, then I'd suppose replacing `!` with a more appropriate name as it's much easier to get confused otherwise.  In essence, either !(x == .) must evaluate to (x != .) if the underlying meaning of these are the same, or the `!` in `!(x==.)` must be replaced to something that's more appropriate for what it's supposed to be. Personally, I prefer the former. It would greatly tighten the structure and consistency. "na.rm = TRUE/FALSE" sounds good to me.  I'd only considered nomatch before in the context of joins, not logical subsets. Yes, I find this option would give more control in evaluating expressions with ease in `i`, by providing both "subset" (default) and the typical data.frame subsetting (na.rm = FALSE). Best regards,   Arun     _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help _______________________________________________datatable-help mailing list[hidden email]https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help _______________________________________________ datatable-help mailing list [hidden email] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

Open this post in threaded view
|
Report Content as Inappropriate

## Re: Follow-up on subsetting data.table with NAs

Open this post in threaded view
|
Report Content as Inappropriate