usually in such cases symlink is made to the actual file / directory.
approximately like this:
$ d=$(date +'%Y/%m/%d') $ mkdir -p $d $ ln -snf $d actual
and it turns out this picture:
$ tree . ├── 2015 │ └── 11 │ └── 03 └── actual -> 2015/11/03 4 directories, 0 files
now by the name of actual (arbitrary) you can refer to the contents of the catalog 2015/11/03 .
and tomorrow, after executing the same commands, symlink actual will already point to the catalog 2015/11/04 .
as applied to your particular case, as I understand it, it should look something like this (instead of actual I used the name last ):
$ d=$(date +'%Y/%m/%d') $ hdfs dfs -mkdir -p user/result/$d $ ln -snf user/result/$d user/result/last $ hdfs dfs -put table.csv user/result/last/
update
I read the comments to this answer - yes, in hdfs there are no symlink- s.
then it probably makes sense to save the required path in the file. or only locally, or (in addition) in the same place, in hdfs . something like this:
$ d=$(date +'%Y/%m/%d') $ hdfs dfs -mkdir -p user/result/$d # сохраняем «переменную» часть пути в локальный файл last $ echo $d > last $ hdfs dfs -put table.csv user/result/$d/ # сохраняем локальный файл last в hdfs $ hdfs dfs -put last user/result/
if you need to get a saved file, do get :
$ hdfs dfs -get user/result/last last
and read the contents of the last file into a variable:
$ d=$(cat last)
which can then be used for addressing:
$ hdfs dfs -get user/result/$d/table.csv table.csv
