By default, make query column names unique (#254)

* By default, make query column names unique

Fixes #253. Most data formats have requirements around column name
uniqueness, including DataFrames.jl. The proposed changes here ensure
query result columns are unique by default, while still allowing column
names to be duplicated if they pass `allowduplicates=true` to
`DBInterface.execute`.

* Update test/runtests.jl

Co-authored-by: Bogumił Kamiński <[email protected]>

Co-authored-by: Bogumił Kamiński <[email protected]>

Author: quinnj

src/SQLite.jl

@@ -458,7 +458,7 @@ end
 execute(stmt::Stmt, params::DBInterface.StatementParams) =
     execute(stmt.db, _stmt(stmt), params)
 
-execute(stmt::Stmt; kwargs...) = execute(stmt, kwargs.data)
+execute(stmt::Stmt; kwargs...) = execute(stmt, values(kwargs))
 
 function execute(db::DB, sql::AbstractString, params::DBInterface.StatementParams)
      # prepare without registering _Stmt in DB
@@ -470,7 +470,7 @@ function execute(db::DB, sql::AbstractString, params::DBInterface.StatementParam
     end
 end
 
-execute(db::DB, sql::AbstractString; kwargs...) = execute(db, sql, kwargs.data)
+execute(db::DB, sql::AbstractString; kwargs...) = execute(db, sql, values(kwargs))
 
 """
     SQLite.esc_id(x::Union{AbstractString,Vector{AbstractString}})

src/tables.jl

@@ -109,14 +109,24 @@ Calling `SQLite.reset!(result)` will re-execute the query and reset the iterator
 The resultset iterator supports the [Tables.jl](https://github.com/JuliaData/Tables.jl) interface, so results can be collected in any Tables.jl-compatible sink,
 like `DataFrame(results)`, `CSV.write("results.csv", results)`, etc.
 """
-function DBInterface.execute(stmt::Stmt, params::DBInterface.StatementParams)
+function DBInterface.execute(stmt::Stmt, params::DBInterface.StatementParams; allowduplicates::Bool=false)
     status = execute(stmt, params)
     _st = _stmt(stmt)
     cols = sqlite3_column_count(_st.handle)
     header = Vector{Symbol}(undef, cols)
     types = Vector{Type}(undef, cols)
     for i = 1:cols
-        header[i] = sym(sqlite3_column_name(_st.handle, i))
+        nm = sym(sqlite3_column_name(_st.handle, i))
+        if !allowduplicates && nm in view(header, 1:(i - 1))
+            j = 1
+            newnm = Symbol(nm, :_, j)
+            while newnm in view(header, 1:(i - 1))
+                j += 1
+                newnm = Symbol(nm, :_, j)
+            end
+            nm = newnm
+        end
+        header[i] = nm
         types[i] = Union{juliatype(_st.handle, i), Missing}
     end
     return Query(stmt, Ref(status), header, types, Dict(x=>i for (i, x) in enumerate(header)))

test/runtests.jl

@@ -424,6 +424,11 @@ tbl3 = (c = [7, 8, 9], a = [4, 5, 6])
 # Test busy_timeout
 @test SQLite.busy_timeout(db, 300) == 0
 
+# 253, ensure query column names are unique by default
+db = SQLite.DB()
+res = DBInterface.execute(db, "select 1 as x2, 2 as x2, 3 as x2, 4 as x2_2") |> columntable
+@test res == (x2 = [1], x2_1 = [2], x2_2 = [3], x2_2_1 = [4])
+
 @testset "load!()/drop!() table name escaping" begin
     tbl = (a = [1, 2, 3], b = ["a", "b", "c"])
     SQLite.load!(tbl, db, "escape 10.0%")