Pin ReadStat_jll to 1.1.9, raise Julia floor to 1.10, fix wrapper defects

Project.toml

@@ -1,6 +1,6 @@
 name = "ReadStat"
 uuid = "d71aba96-b539-5138-91ee-935c3ee1374c"
-version = "1.1.2-DEV"
+version = "1.2.0-DEV"
 
 [deps]
 Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
@@ -12,9 +12,9 @@ TestItemRunner = "f8b46487-2199-4994-9208-9a1283c18c0a"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [compat]
-julia = "1.3"
+julia = "1.10"
 DataValues = "0.4.13, 0.5, 1"
-ReadStat_jll = "1.1.1"
+ReadStat_jll = "1.1.9"
 
 [targets]
 test = ["Test", "TestItemRunner"]

src/C_interface.jl

@@ -4,7 +4,7 @@
 end
 
 function readstat_get_modified_time(metadata::Ptr{Nothing})
-    return ccall((:readstat_get_modified_time, libreadstat), UInt, (Ptr{Nothing},), metadata)
+    return ccall((:readstat_get_modified_time, libreadstat), Int64, (Ptr{Nothing},), metadata)
 end
 
 function readstat_get_file_format_version(metadata::Ptr{Nothing})
@@ -55,23 +55,23 @@
     return ccall((:readstat_value_type, libreadstat), Cint, (Value,), val)
 end
 
 function readstat_parse(filename::String, type::Val{:dta}, parser::Ptr{Nothing}, ds::ReadStatDataFrame)
     return ccall((:readstat_parse_dta, libreadstat), Cint, (Ptr{Nothing}, Cstring, Any), parser, string(filename), ds)
 end
 
 function readstat_parse(filename::String, type::Val{:sav}, parser::Ptr{Nothing}, ds::ReadStatDataFrame)
     return ccall((:readstat_parse_sav, libreadstat), Cint, (Ptr{Nothing}, Cstring, Any), parser, string(filename), ds)
 end
 
 function readstat_parse(filename::String, type::Val{:por}, parser::Ptr{Nothing}, ds::ReadStatDataFrame)
     return ccall((:readstat_parse_por, libreadstat), Cint, (Ptr{Nothing}, Cstring, Any), parser, string(filename), ds)
 end
 
 function readstat_parse(filename::String, type::Val{:sas7bdat}, parser::Ptr{Nothing}, ds::ReadStatDataFrame)
     return ccall((:readstat_parse_sas7bdat, libreadstat), Cint, (Ptr{Nothing}, Cstring, Any), parser, string(filename), ds)
 end
 
 function readstat_parse(filename::String, type::Val{:xport}, parser::Ptr{Nothing}, ds::ReadStatDataFrame)
     return ccall((:readstat_parse_xport, libreadstat), Cint, (Ptr{Nothing}, Cstring, Any), parser, string(filename), ds)
 end
 

src/ReadStat.jl

@@ -21,12 +21,17 @@
 ##############################################################################
 
 const READSTAT_TYPE_STRING      = Cint(0)
-const READSTAT_TYPE_CHAR        = Cint(1)
+const READSTAT_TYPE_INT8        = Cint(1)
 const READSTAT_TYPE_INT16       = Cint(2)
 const READSTAT_TYPE_INT32       = Cint(3)
 const READSTAT_TYPE_FLOAT       = Cint(4)
 const READSTAT_TYPE_DOUBLE      = Cint(5)
-const READSTAT_TYPE_LONG_STRING = Cint(6)
+const READSTAT_TYPE_STRING_REF  = Cint(6)
+
+# Julia type for each readstat_type_t, indexed by the enum value plus one.
+# STRING_REF is a reference into a string table, so it surfaces as a String
+# just like STRING does.
+const READSTAT_TYPES = (String, Int8, Int16, Int32, Float32, Float64, String)
 
 const READSTAT_ERROR_OPEN       = Cint(1)
 const READSTAT_ERROR_READ       = Cint(2)
@@ -85,13 +90,6 @@
 ##
 ##############################################################################
 
-function handle_info!(obs_count::Cint, var_count::Cint, ds_ptr::Ptr{ReadStatDataFrame})
-    ds = unsafe_pointer_to_objref(ds_ptr)
-    ds.rows = obs_count
-    ds.columns = var_count
-    return Cint(0)
-end
-
 function handle_metadata!(metadata::Ptr{Nothing}, ds_ptr::Ptr{ReadStatDataFrame})
     ds = unsafe_pointer_to_objref(ds_ptr)
     ds.filelabel = readstat_get_file_label(metadata)
@@ -114,33 +112,18 @@
     ptr == C_NULL ? "" : unsafe_string(ptr)
 end
 
-function get_type(data_type::Cint)
-    if data_type == READSTAT_TYPE_STRING
-        return String
-    elseif data_type == READSTAT_TYPE_CHAR
-        return Int8
-    elseif data_type == READSTAT_TYPE_INT16
-        return Int16
-    elseif data_type == READSTAT_TYPE_INT32
-        return Int32
-    elseif data_type == READSTAT_TYPE_FLOAT
-        return Float32
-    elseif data_type == READSTAT_TYPE_DOUBLE
-        return Float64
-    end
-    return Nothing
-end
+get_type(data_type::Cint) = READSTAT_TYPES[data_type + 1]
 get_type(variable::Ptr{Nothing}) = get_type(readstat_variable_get_type(variable))
 
 get_storagewidth(variable::Ptr{Nothing}) = readstat_variable_get_storage_width(variable)
 
 get_measure(variable::Ptr{Nothing}) = readstat_variable_get_measure(variable)
 
-get_alignment(variable::Ptr{Nothing}) = readstat_variable_get_measure(variable)
+get_alignment(variable::Ptr{Nothing}) = readstat_variable_get_alignment(variable)
 
 function handle_variable!(var_index::Cint, variable::Ptr{Nothing},
                          val_label::Cstring,  ds_ptr::Ptr{ReadStatDataFrame})
     col = var_index + 1
     ds = unsafe_pointer_to_objref(ds_ptr)::ReadStatDataFrame
     missing_count = readstat_variable_get_missing_ranges_count(variable)
 
@@ -165,10 +148,7 @@
     return Cint(0)
 end
 
-function get_type(val::Value)
-    data_type = readstat_value_type(val)
-    return [String, Int8, Int16, Int32, Float32, Float64, String][data_type + 1]
-end
+get_type(val::Value) = get_type(readstat_value_type(val))
 
 Base.convert(::Type{Int8}, val::Value) = ccall((:readstat_int8_value, libreadstat), Int8, (Value,), val)
 Base.convert(::Type{Int16}, val::Value) = ccall((:readstat_int16_value, libreadstat), Int16, (Value,), val)
@@ -265,15 +245,14 @@
 
 function Parser()
     parser = ccall((:readstat_parser_init, libreadstat), Ptr{Nothing}, ())
-    info_fxn = @cfunction(handle_info!, Cint, (Cint, Cint, Ptr{ReadStatDataFrame}))
     meta_fxn = @cfunction(handle_metadata!, Cint, (Ptr{Nothing}, Ptr{ReadStatDataFrame}))
     var_fxn = @cfunction(handle_variable!, Cint, (Cint, Ptr{Nothing}, Cstring,  Ptr{ReadStatDataFrame}))
     val_fxn = @cfunction(handle_value!, Cint, (Cint, Ptr{Nothing}, ReadStatValue, Ptr{ReadStatDataFrame}))
     label_fxn = @cfunction(handle_value_label!, Cint, (Cstring, Value, Cstring, Ptr{ReadStatDataFrame}))
-    ccall((:readstat_set_metadata_handler, libreadstat), Int, (Ptr{Nothing}, Ptr{Nothing}), parser, meta_fxn)
-    ccall((:readstat_set_variable_handler, libreadstat), Int, (Ptr{Nothing}, Ptr{Nothing}), parser, var_fxn)
-    ccall((:readstat_set_value_handler, libreadstat), Int, (Ptr{Nothing}, Ptr{Nothing}), parser, val_fxn)
-    ccall((:readstat_set_value_label_handler, libreadstat), Int, (Ptr{Nothing}, Ptr{Nothing}), parser, label_fxn)
+    ccall((:readstat_set_metadata_handler, libreadstat), Cint, (Ptr{Nothing}, Ptr{Nothing}), parser, meta_fxn)
+    ccall((:readstat_set_variable_handler, libreadstat), Cint, (Ptr{Nothing}, Ptr{Nothing}), parser, var_fxn)
+    ccall((:readstat_set_value_handler, libreadstat), Cint, (Ptr{Nothing}, Ptr{Nothing}), parser, val_fxn)
+    ccall((:readstat_set_value_label_handler, libreadstat), Cint, (Ptr{Nothing}, Ptr{Nothing}), parser, label_fxn)
     return parser
 end
 

test/test_readstat.jl

@@ -1,11 +1,13 @@
 @testitem "ReadStat" begin
     using DataValues
 
-    @testset "ReadStat: $ext files" for (reader, ext) in
-        ((read_dta, "dta"),
-         (read_sav, "sav"),
-         (read_sas7bdat, "sas7bdat"),
-         (read_xport, "xpt"))
+    # Expected alignments are readstat_alignment_t values:
+    # 0 = UNKNOWN, 1 = LEFT, 2 = CENTER, 3 = RIGHT.
+    @testset "ReadStat: $ext files" for (reader, ext, alignments) in
+        ((read_dta, "dta", Int32[3, 3, 3, 3, 3, 3]),
+         (read_sav, "sav", Int32[0, 0, 0, 0, 0, 0]),
+         (read_sas7bdat, "sas7bdat", Int32[0, 0, 0, 0, 0, 0]),
+         (read_xport, "xpt", Int32[3, 3, 3, 3, 3, 1]))
 
         dtafile = joinpath(@__DIR__, "types.$ext")
         rsdf = reader(dtafile)
@@ -19,5 +21,15 @@
         @test data[4] == DataValueArray{Int16}([2, 7, NA])
         @test data[5] == DataValueArray{Int8}([2, 7., NA])
         @test data[6] == DataValueArray{String}(["2", "7", ""])
+
+        # Alignments must come from readstat_variable_get_alignment, not from
+        # the measure accessor sitting next to it in the C API. Every fixture
+        # reports measure UNKNOWN, so reading the wrong one yields all zeros.
+        @test rsdf.alignments == alignments
+        @test rsdf.measures == Int32[0, 0, 0, 0, 0, 0]
+
+        # Every readstat_type_t the readers emit must map to a concrete Julia
+        # type; a gap in that mapping used to surface as Nothing.
+        @test all(!=(Nothing), rsdf.types)
     end
 end