fix: use decimal notation for float formatting to match ODBC sqlcmd

pkg/sqlcmd/format.go

@@ -9,6 +9,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"strconv"
 	"strings"
 	"time"
 
@@ -59,6 +60,11 @@ const (
 	ControlReplaceConsecutive
 )
 
+const (
+	realDefaultWidth  int64 = 14 // REAL and SMALLMONEY
+	floatDefaultWidth int64 = 24 // FLOAT and MONEY
+)
+
 type columnDetail struct {
 	displayWidth       int64
 	leftJustify        bool
@@ -371,11 +377,11 @@ func calcColumnDetails(cols []*sql.ColumnType, fixed int64, variable int64) ([]c
 			columnDetails[i].displayWidth = max64(21, nameLen)
 		case "REAL", "SMALLMONEY":
 			columnDetails[i].leftJustify = false
-			columnDetails[i].displayWidth = max64(14, nameLen)
+			columnDetails[i].displayWidth = max64(realDefaultWidth, nameLen)
 			columnDetails[i].zeroesAfterDecimal = true
 		case "FLOAT", "MONEY":
 			columnDetails[i].leftJustify = false
-			columnDetails[i].displayWidth = max64(24, nameLen)
+			columnDetails[i].displayWidth = max64(floatDefaultWidth, nameLen)
 			columnDetails[i].zeroesAfterDecimal = true
 		case "DECIMAL":
 			columnDetails[i].leftJustify = false
@@ -530,6 +536,10 @@ func (f *sqlCmdFormatterType) scanRow(rows *sql.Rows) ([]string, error) {
 				} else {
 					row[n] = "0"
 				}
+			case float64:
+				row[n] = formatFloat(x, 64, f.columnDetails[n])
+			case float32:
+				row[n] = formatFloat(float64(x), 32, f.columnDetails[n])
 			default:
 				var err error
 				if row[n], err = fmt.Sprintf("%v", x), nil; err != nil {
@@ -552,6 +562,28 @@ func dateTimeFormatString(scale int, addOffset bool) string {
 	return format
 }
 
+// formatFloat formats a float value to match ODBC sqlcmd behavior.
+// Uses decimal notation for typical values, falls back to scientific for extreme values.
+func formatFloat(x float64, bitSize int, col columnDetail) string {
+	formatted := strconv.FormatFloat(x, 'f', -1, bitSize)
+
+	// Determine width threshold for fallback to scientific notation
+	threshold := col.displayWidth
+	if threshold == 0 {
+		typeName := col.col.DatabaseTypeName()
+		if typeName == "REAL" || typeName == "SMALLMONEY" {
+			threshold = realDefaultWidth
+		} else {
+			threshold = floatDefaultWidth
+		}
+	}
+
+	if int64(len(formatted)) > threshold {
+		formatted = strconv.FormatFloat(x, 'g', -1, bitSize)
+	}
+	return formatted
+}
+
 // Prints the final version of a cell based on formatting variables and command line parameters
 func (f *sqlCmdFormatterType) printColumnValue(val string, col int) {
 	c := f.columnDetails[col]

pkg/sqlcmd/format_test.go

@@ -162,3 +162,30 @@ func TestFormatterXmlMode(t *testing.T) {
 	assert.NoError(t, err, "runSqlCmd returned error")
 	assert.Equal(t, `<sys.databases name="master"/>`+SqlcmdEol, buf.buf.String())
 }
+
+func TestFormatterFloatDecimalNotation(t *testing.T) {
+	s, buf := setupSqlCmdWithMemoryOutput(t)
+	defer func() { _ = buf.Close() }()
+
+	s.vars.Set(SQLCMDMAXVARTYPEWIDTH, "256")
+	query := `SELECT CAST(4713347.3103808956 AS FLOAT) as val`
+	err := runSqlCmd(t, s, []string{query, "GO"})
+	assert.NoError(t, err)
+
+	output := buf.buf.String()
+	assert.NotContains(t, output, "e+", "typical floats should use decimal notation")
+	assert.Contains(t, output, "4713347.310380", "should contain decimal value")
+}
+
+func TestFormatterFloatScientificFallback(t *testing.T) {
+	s, buf := setupSqlCmdWithMemoryOutput(t)
+	defer func() { _ = buf.Close() }()
+
+	s.vars.Set(SQLCMDMAXVARTYPEWIDTH, "256")
+	query := `SELECT CAST(1e100 AS FLOAT) as val`
+	err := runSqlCmd(t, s, []string{query, "GO"})
+	assert.NoError(t, err)
+
+	output := buf.buf.String()
+	assert.Contains(t, output, "e+", "extreme values should use scientific notation")
+}