This is a very, VERY old answer, but I see I started a draft.
At 12:12 AM 8/19/2006, Dan Olsen wrote:
>I am working on a project which requires import of data into SPSS from
>a file created by a Web Application. The application itself already is
>setup to write out flat files (.xls,.tab,.csv). At present, the data
>can be imported into SPSS just fine, but then the SPSS Analyst has to
>create the Data definitions and any Variable types they need. [I want
>to] map our data types to their functional analogs in SPSS so they
>show up in the 'Variable View' tab with the correct restrictions.
>
>my variable types:
>text : alphanumeric string
>integer : max/min
>float : max/min
>enum : set with possible values and indices
>bool : 1:0
>
>i have seen some syntax that indicates i can use 'DATA LIST' to do
>some type definition, but i don't have the exact syntax for my
>variable types. also, i would prefer not to have to use a fixed-width
>definition.
Did you get this resolved, or give up, or is it still open?
DATA LIST is not at all restricted to fixed-width *input*. SPSS string
variables are fixed-width; that's simply a design limitation. The
standard practice is to declare the strings to cover the greatest width
you'll need, and live with the wasted space.
>At present, the data can be imported into SPSS just fine, but then the
>SPSS Analyst has to create the Data definitions and any Variable types
>they need.
To assign variable names, either you use field names you can read from
your data; or, somebody puts them in by hand; or, you generate
arbitrary ones like VAR001, VAR002, etc. (The last is actually a fairly
good choice if there are many fields, but I'd recommend a prefix that
says something about the data source or project, rather than "VAR".)
>my variable types:
>text : alphanumeric string
>integer : max/min
>float : max/min
>enum : set with possible values and indices
>bool : 1:0
Now, SPSS has only two datatypes: fixed-length alphanumeric string, and
numeric. (Numeric are stored as 64-bit floating.)
All your numeric types (I'm including 'enum' as numeric, though it can
be mapped into string variables as well) can be represented nicely as
SPSS numeric, often with wasted space but what the heck.
What SPSS *can't* do is enforce the value constraints; or even check
them, during input.
If the variable definitions, including types and constraints, are in a
form that's reasonably easy to parse, and if there are a lot of
variables, I'd probably write SPSS code that parsed the definitions and
generated SPSS code to apply appropriate formats, variable and value
labels, and syntax to check the constraints. If you have SPSS 14+, and
know or want to learn Python, the parsing and code-generating may be
easier in Python. And the constraints can be made user-defined variable
attributes, so Python can look them up and generate the code to check
them, directly from the SPSS file.
Locked
