Le dimanche 20 mars 2016 à 16:56 -0700, Jeffrey Sarnoff a écrit : > Redis itself is written in C. They document GET key: > > Get the value of key. If the key does not exist the special > > value nil is returned. > > An error is returned if the value stored at key is not a string, > because GET only handles string values. > > digging deeper > > The client library API should not return an empty string, but a nil > > object, when the server replies with a Null Bulk String. > > For example a Ruby library should return 'nil' while a C library > should return NULL (or set a special flag in the reply object), and > so forth. > > Single elements of an Array may be Null. This is used in Redis > > replies in order to signal that this elements are missing and not > > empty strings. > > This can happen with the SORT command when used with the > GET pattern option when the specified key is missing. > > [For example, if the] second element is a Null. The client library > should return something like this: ["foo", nil, "bar"] > The Redis nil indicates a non-present value (missing or unavailable > or, not extant: a domain|-> range error). > From a semiotic viewpoint, Julia's nothing is closer to "absence" > than it is to "an absent value"; of course > the operational machinery supplies an actual entity to be nothing (a > singleton realization ofthe type Void). > > A much more contextual fit be the use of Nullable, although that may > require more of the client; a simpler way > to handle the Redis nil without doing something with nothing is to > use a dedicated const symbol or a singleton > to be that sentinel, perhaps: > const RedisNil = :RedisNil # or > type RedisNIL end; RedisNil = > RedisNIL() > > Somewhere there are lengthy and informative discussions about Julia > and nil / NULL / nothing. > (I noted this thread to the Redis.jl project). Interesting. Though this wouldn't be an actual sentinel, as a symbol is a different type from a string, so type instability would remain.
It seems that Redis.jl has chosen to follow the Python API, which isn't type-stable. In Julia, it could make sense to retain a type-stable solution, i.e. always return a Nullable. This would be a good test to check what improvements Julia needs to make working with Nullable pleasant enough. Actually, get() from Redis.jl is faced with the very same design issue as get() from Julia Base. The latter raises an error when a key is missing, and there's been discussions about offering an alternative function which would return a Nullable: https://github.com/JuliaLang/julia/issues/13055 I would argue that Redis.jl should follow the same pattern as Base for consistency. It would be interesting to get comments from the package author about this. Regards > > > > > > Redis.jl returns nothing when requesting a the value of a key that > > doesn't exist: > > > > using Redis > > conn = RedisConnection() > > r = get(conn, "non_existent_key") > > disconnect(conn) > > r == nothing # true > > > > > > > For now I don't know of a good solution to this pattern, but > > > there's > > > been some discussion about it: > > > https://github.com/JuliaLang/julia/issues/15174 > > > > > > You should definitely use a Nullable instead of returning > > > nothing. > > > > > > > > > Regards > > > > > > Le samedi 19 mars 2016 à 02:58 -0700, Jeffrey Sarnoff a écrit : > > > > You may be misusing nothing. It is unusual that a function > > > would > > > > return nothing some of the time and something other times. > > > > Take a look at http://docs.julialang.org/en/latest/manual/faq/# > > > nothin > > > > gness-and-missing-values > > > > If you have additional questions about this, please give an > > > example > > > > of what get_a(...) is getting and why it would be nothing some > > > of the > > > > time. > > > > > > > > > Hi All > > > > > > > > > > > > > > > I found my self writing code like this a lot: > > > > > > > > > > x = get_a(...) > > > > > > > > > > if x != nothing > > > > > y::A = x > > > > > do_sth(y, ...) > > > > > end > > > > > > > > > > In the above, I have to check for nothing first, and if it is > > > not > > > > > nothing, then I do a type assert to make sure the type is > > > what I > > > > > expected. > > > > > > > > > > Is there any function or macro in Julia that can help this? > > > > > > > > > > I know in F#, I have option.bind, so option.bind f x is > > > equivalent > > > > > to a pattern match: if x is None - > None; if x is something > > > -> > > > > > f(something) > > > > > > > > > > Also in C#, I have "customers?[0]?.Orders?.Count();" (as > > > long as > > > > > there is null before ?, it returns null immediately) > > > > > > > > > > Does Julia have something similar? > > > > > > > > > >
