Object Design for Metadata and Results

Data Structures

Motivation

See issues

• SASDataResult object
• read function should be more lean

ResultSet

ResultSet is the primary object that represents data returned from reading a SAS data file. ResultSet implements the Base.Iteration interface as well as the IterableTables.jl interface.

Fields

• columns: a vector of columns, each being a vector itself
• names: a vector of column symbols
• size: a tuple (nrows, ncols)

Accessors

• columns(::ResultSet)
• names(::ResultSet)
• size(::ResultSet)
• size(::ResultSet, dim::Integer)

Single Row/Column Indexing

• rs[i] returns a tuple for row i
• rs[:c] returns a vector for column c

Multiple Row/Column Indexing

• rs[i:j] returns a view of ResultSet with rows between i and j
• rs[:c...] returns a view of ResultSet with columns specified

Cell Indexing

• rs[i,j] returns a single value for row i column j
• rs[i,:c] returns a single value for row i column c e.g. rs[:A, :B]
• Specific cell can be assigned using the above indexing methods

Usage

rs = readsas("abc.sas7bdat")
df = DataFrame(columns(rs), names(rs))
size(rs)

Note: A design decision is made here to NOT include file-specific meta data in the result set.

Metadata

Metadata contains information about a SAS data file.

Fields

• filename: file name/path of the SAS data set
• encoding: file encoding e.g. "ISO8859-1"
• endianness: either :LittleEndian`` or :BigEndian`
• compression: could be :RLE, :RDC, or :none
• pagesize: size of each data page in bytes
• npages: number of pages in the file
• nrows: number of data rows in the file
• ncols: number of data columns in the file
• columnsinfo: vector of column symbols and their respective types (Float64 or String)

Usage

metadata("abc.sas7bdat")

handler = SASLib.open("abc.sas7bdat")
metadata(handler)

Note: Direct field access is encouraged for metadata.

Interface Changes

readsas returns ::ResultSet

read returns ::ResultSet

metadata (a new function) returns ::Metadata