How to homogenize NA values within a .gpkg from R

Question

I have created an SpatialPointsDataFrame in R with some data. Here is the outputs from R:

> class(Fotos)
[1] "SpatialPointsDataFrame"
attr(,"package")
[1] "sp"

> class(Fotos$Azimuth)
[1] "numeric"

> Fotos$Azimuth
  [1]  NA  NA  NA  NA   0  NA  NA  NA  NA  NA
 [11]  45  NA  90  NA  NA  NA 135  NA  NA  NA
 [21]  NA 180  NA  NA  NA  NA  NA  NA 225  NA
 [31]  NA  NA  NA  NA 270 315  NA  NA  NA  NA
 ...

> class(Fotos$labelpos)
[1] "numeric"

> Fotos$labelpos
  [1] NA NA NA NA  0 NA NA NA NA NA  0 NA  0 NA
 [15] NA NA  0 NA NA NA NA  0 NA NA NA NA NA NA
 [29]  0 NA NA NA NA NA  0  1 NA NA NA NA NA NA
 [43] NA NA NA NA NA NA  0 NA NA NA NA NA  1 NA
 ...

It seems that both Attributes have the same data structure and class but when I export this SPDF to .gpkg some attributes replace NA by NULL and others replace NA by nothing. Like here:

How could I homogeneize those null/na/empty values in the .gpkg exported from R? I tryed something like this without fourtune:

Fotos@data[is.na(Fotos@data)] <- NA

EDIT: I tried to homogenize those NA with this:

 # Homonegeneizar NAs
    storage.mode(Fotos$OrdenSel) <- "integer"
    storage.mode(Fotos$Plano) <- "integer"
    storage.mode(Fotos$Cluster) <- "integer"
    storage.mode(Fotos$labelpos) <- "integer"
    storage.mode(Fotos$Azimuth) <- "integer"

But QGIS read those values as integers, but some has NULL and others empty...

Answer 1

If I create a test spatial data frame and write to a GeoPackage:

 Fotos = data.frame(x=1:10,y=1:10, Azimuth=sample(180,10), labelpos=sample(c(0,1),10,TRUE))
 Fotos$labelpos[c(2,5,6)]=NA
 Fotos$Azimuth[c(2,5,6)]=NA
 coordinates(Fotos)=~x+y
 writeOGR(Fotos,"Test.gpkg",layer="Fotos",driver="GPKG")

Then I see that QGIS has interpreted the Azimuth column as "int" and the labelpos column as "double":

The missing values are then shown as you see them:

So I think its a difference in the way QGIS displays missing values in integer or real attribute columns. For an int it says NULL, and for a double it shows blank.

Looking at the storage mode of each column reveals why QGIS thinks this way:

> storage.mode(Fotos$labelpos)
[1] "double"
> storage.mode(Fotos$Azimuth)
[1] "integer"

that has gone through into the schema in the GeoPackage:

CREATE TABLE "Fotos" ( 
     "fid" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL, 
     "geom" POINT,
     "Azimuth" MEDIUMINT,
     "labelpos" REAL);

Quite why R creates them with those storage modes, or why QGIS uses different text for missing data in ints and reals is another question...

But there is clearly no reason to try and clean up your data frame - its as good as it gets.