// Copyright 2016 - 2019 The excelize Authors. All rights reserved. Use of
// this source code is governed by a BSD-style license that can be found in
// the LICENSE file.
//
// Package excelize providing a set of functions that allow you to write to
// and read from XLSX files. Support reads and writes XLSX file generated by
// Microsoft Excelâ„¢ 2007 and later. Support save file without losing original
// charts of XLSX. This library needs Go version 1.8 or later.
package excelize
import (
"bytes"
"encoding/xml"
"fmt"
"io"
"math"
"strconv"
)
// GetRows return all the rows in a sheet by given worksheet name (case
// sensitive). For example:
//
// for _, row := range xlsx.GetRows("Sheet1") {
// for _, colCell := range row {
// fmt.Print(colCell, "\t")
// }
// fmt.Println()
// }
//
func (f *File) GetRows(sheet string) [][]string {
name, ok := f.sheetMap[trimSheetName(sheet)]
if !ok {
return nil
}
xlsx := f.workSheetReader(sheet)
if xlsx != nil {
output, _ := xml.Marshal(f.Sheet[name])
f.saveFileList(name, replaceWorkSheetsRelationshipsNameSpaceBytes(output))
}
xml.NewDecoder(bytes.NewReader(f.readXML(name)))
d := f.sharedStringsReader()
var (
inElement string
rowData xlsxRow
)
rowCount, colCount := f.getTotalRowsCols(name)
rows := make([][]string, rowCount)
for i := range rows {
rows[i] = make([]string, colCount+1)
}
var row int
decoder := xml.NewDecoder(bytes.NewReader(f.readXML(name)))
for {
token, _ := decoder.Token()
if token == nil {
break
}
switch startElement := token.(type) {
case xml.StartElement:
inElement = startElement.Name.Local
if inElement == "row" {
rowData = xlsxRow{}
_ = decoder.DecodeElement(&rowData, &startElement)
cr := rowData.R - 1
for _, colCell := range rowData.C {
col, _ := MustCellNameToCoordinates(colCell.R)
val, _ := colCell.getValueFrom(f, d)
rows[cr][col-1] = val
if val != "" {
row = rowData.R
}
}
}
default:
}
}
return rows[:row]
}
// Rows defines an iterator to a sheet
type Rows struct {
decoder *xml.Decoder
token xml.Token
err error
f *File
}
// Next will return true if find the next row element.
func (rows *Rows) Next() bool {
for {
rows.token, rows.err = rows.decoder.Token()
if rows.err == io.EOF {
rows.err = nil
}
if rows.token == nil {
return false
}
switch startElement := rows.token.(type) {
case xml.StartElement:
inElement := startElement.Name.Local
if inElement == "row" {
return true
}
}
}
}
// Error will return the error when the find next row element
func (rows *Rows) Error() error {
return rows.err
}
// Columns return the current row's column values
func (rows *Rows) Columns() []string {
if rows.token == nil {
return []string{}
}
startElement := rows.token.(xml.StartElement)
r := xlsxRow{}
_ = rows.decoder.DecodeElement(&r, &startElement)
d := rows.f.sharedStringsReader()
columns := make([]string, len(r.C))
for _, colCell := range r.C {
col, _ := MustCellNameToCoordinates(colCell.R)
val, _ := colCell.getValueFrom(rows.f, d)
columns[col-1] = val
}
return columns
}
// ErrSheetNotExist defines an error of sheet is not exist
type ErrSheetNotExist struct {
SheetName string
}
func (err ErrSheetNotExist) Error() string {
return fmt.Sprintf("Sheet %s is not exist", string(err.SheetName))
}
// Rows return a rows iterator. For example:
//
// rows, err := xlsx.Rows("Sheet1")
// for rows.Next() {
// for _, colCell := range rows.Columns() {
// fmt.Print(colCell, "\t")
// }
// fmt.Println()
// }
//
func (f *File) Rows(sheet string) (*Rows, error) {
xlsx := f.workSheetReader(sheet)
name, ok := f.sheetMap[trimSheetName(sheet)]
if !ok {
return nil, ErrSheetNotExist{sheet}
}
if xlsx != nil {
output, _ := xml.Marshal(f.Sheet[name])
f.saveFileList(name, replaceWorkSheetsRelationshipsNameSpaceBytes(output))
}
return &Rows{
f: f,
decoder: xml.NewDecoder(bytes.NewReader(f.readXML(name))),
}, nil
}
// getTotalRowsCols provides a function to get total columns and rows in a
// worksheet.
func (f *File) getTotalRowsCols(name string) (int, int) {
decoder := xml.NewDecoder(bytes.NewReader(f.readXML(name)))
var inElement string
var r xlsxRow
var tr, tc int
for {
token, _ := decoder.Token()
if token == nil {
break
}
switch startElement := token.(type) {
case xml.StartElement:
inElement = startElement.Name.Local
if inElement == "row" {
r = xlsxRow{}
_ = decoder.DecodeElement(&r, &startElement)
tr = r.R
for _, colCell := range r.C {
col, _ := MustCellNameToCoordinates(colCell.R)
if col > tc {
tc = col
}
}
}
default:
}
}
return tr, tc
}
// SetRowHeight provides a function to set the height of a single row. For
// example, set the height of the first row in Sheet1:
//
// xlsx.SetRowHeight("Sheet1", 1, 50)
//
func (f *File) SetRowHeight(sheet string, row int, height float64) {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
xlsx := f.workSheetReader(sheet)
prepareSheetXML(xlsx, 0, row)
rowIdx := row - 1
xlsx.SheetData.Row[rowIdx].Ht = height
xlsx.SheetData.Row[rowIdx].CustomHeight = true
}
// getRowHeight provides a function to get row height in pixels by given sheet
// name and row index.
func (f *File) getRowHeight(sheet string, row int) int {
xlsx := f.workSheetReader(sheet)
for _, v := range xlsx.SheetData.Row {
if v.R == row+1 && v.Ht != 0 {
return int(convertRowHeightToPixels(v.Ht))
}
}
// Optimisation for when the row heights haven't changed.
return int(defaultRowHeightPixels)
}
// GetRowHeight provides a function to get row height by given worksheet name
// and row index. For example, get the height of the first row in Sheet1:
//
// xlsx.GetRowHeight("Sheet1", 1)
//
func (f *File) GetRowHeight(sheet string, row int) float64 {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
xlsx := f.workSheetReader(sheet)
if row > len(xlsx.SheetData.Row) {
return defaultRowHeightPixels // it will be better to use 0, but we take care with BC
}
for _, v := range xlsx.SheetData.Row {
if v.R == row && v.Ht != 0 {
return v.Ht
}
}
// Optimisation for when the row heights haven't changed.
return defaultRowHeightPixels
}
// sharedStringsReader provides a function to get the pointer to the structure
// after deserialization of xl/sharedStrings.xml.
func (f *File) sharedStringsReader() *xlsxSST {
if f.SharedStrings == nil {
var sharedStrings xlsxSST
ss := f.readXML("xl/sharedStrings.xml")
if len(ss) == 0 {
ss = f.readXML("xl/SharedStrings.xml")
}
_ = xml.Unmarshal(namespaceStrictToTransitional(ss), &sharedStrings)
f.SharedStrings = &sharedStrings
}
return f.SharedStrings
}
// getValueFrom return a value from a column/row cell, this function is
// inteded to be used with for range on rows an argument with the xlsx opened
// file.
func (xlsx *xlsxC) getValueFrom(f *File, d *xlsxSST) (string, error) {
switch xlsx.T {
case "s":
xlsxSI := 0
xlsxSI, _ = strconv.Atoi(xlsx.V)
if len(d.SI[xlsxSI].R) > 0 {
value := ""
for _, v := range d.SI[xlsxSI].R {
value += v.T
}
return value, nil
}
return f.formattedValue(xlsx.S, d.SI[xlsxSI].T), nil
case "str":
return f.formattedValue(xlsx.S, xlsx.V), nil
case "inlineStr":
return f.formattedValue(xlsx.S, xlsx.IS.T), nil
default:
return f.formattedValue(xlsx.S, xlsx.V), nil
}
}
// SetRowVisible provides a function to set visible of a single row by given
// worksheet name and Excel row number. For example, hide row 2 in Sheet1:
//
// xlsx.SetRowVisible("Sheet1", 2, false)
//
func (f *File) SetRowVisible(sheet string, row int, visible bool) {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
xlsx := f.workSheetReader(sheet)
prepareSheetXML(xlsx, 0, row)
xlsx.SheetData.Row[row-1].Hidden = !visible
}
// GetRowVisible provides a function to get visible of a single row by given
// worksheet name and Excel row number. For example, get visible state of row
// 2 in Sheet1:
//
// xlsx.GetRowVisible("Sheet1", 2)
//
func (f *File) GetRowVisible(sheet string, row int) bool {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
xlsx := f.workSheetReader(sheet)
if row > len(xlsx.SheetData.Row) {
return false
}
return !xlsx.SheetData.Row[row-1].Hidden
}
// SetRowOutlineLevel provides a function to set outline level number of a
// single row by given worksheet name and Excel row number. For example,
// outline row 2 in Sheet1 to level 1:
//
// xlsx.SetRowOutlineLevel("Sheet1", 2, 1)
//
func (f *File) SetRowOutlineLevel(sheet string, row int, level uint8) {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
xlsx := f.workSheetReader(sheet)
prepareSheetXML(xlsx, 0, row)
xlsx.SheetData.Row[row-1].OutlineLevel = level
}
// GetRowOutlineLevel provides a function to get outline level number of a
// single row by given worksheet name and Excel row number. For example, get
// outline number of row 2 in Sheet1:
//
// xlsx.GetRowOutlineLevel("Sheet1", 2)
//
func (f *File) GetRowOutlineLevel(sheet string, row int) uint8 {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
xlsx := f.workSheetReader(sheet)
if row > len(xlsx.SheetData.Row) {
return 0
}
return xlsx.SheetData.Row[row-1].OutlineLevel
}
// RemoveRow provides a function to remove single row by given worksheet name
// and Excel row number. For example, remove row 3 in Sheet1:
//
// xlsx.RemoveRow("Sheet1", 3)
//
// Use this method with caution, which will affect changes in references such
// as formulas, charts, and so on. If there is any referenced value of the
// worksheet, it will cause a file error when you open it. The excelize only
// partially updates these references currently.
func (f *File) RemoveRow(sheet string, row int) {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
xlsx := f.workSheetReader(sheet)
if row > len(xlsx.SheetData.Row) {
return
}
for rowIdx := range xlsx.SheetData.Row {
if xlsx.SheetData.Row[rowIdx].R == row {
xlsx.SheetData.Row = append(xlsx.SheetData.Row[:rowIdx],
xlsx.SheetData.Row[rowIdx+1:]...)[:len(xlsx.SheetData.Row)-1]
f.adjustHelper(sheet, rows, row, -1)
return
}
}
}
// InsertRow provides a function to insert a new row after given Excel row
// number starting from 1. For example, create a new row before row 3 in
// Sheet1:
//
// xlsx.InsertRow("Sheet1", 3)
//
func (f *File) InsertRow(sheet string, row int) {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
f.adjustHelper(sheet, rows, row, 1)
}
// DuplicateRow inserts a copy of specified row (by it Excel row number) below
//
// xlsx.DuplicateRow("Sheet1", 2)
//
// Use this method with caution, which will affect changes in references such
// as formulas, charts, and so on. If there is any referenced value of the
// worksheet, it will cause a file error when you open it. The excelize only
// partially updates these references currently.
func (f *File) DuplicateRow(sheet string, row int) {
f.DuplicateRowTo(sheet, row, row+1)
}
// DuplicateRowTo inserts a copy of specified row by it Excel number
// to specified row position moving down exists rows after target position
//
// xlsx.DuplicateRowTo("Sheet1", 2, 7)
//
// Use this method with caution, which will affect changes in references such
// as formulas, charts, and so on. If there is any referenced value of the
// worksheet, it will cause a file error when you open it. The excelize only
// partially updates these references currently.
func (f *File) DuplicateRowTo(sheet string, row, row2 int) {
if row < 1 {
panic(newInvalidRowNumberError(row)) // Fail fast to avoid possible future side effects!
}
xlsx := f.workSheetReader(sheet)
if row > len(xlsx.SheetData.Row) || row2 < 1 || row == row2 {
return
}
var ok bool
var rowCopy xlsxRow
for i, r := range xlsx.SheetData.Row {
if r.R == row {
rowCopy = xlsx.SheetData.Row[i]
ok = true
break
}
}
if !ok {
return
}
f.adjustHelper(sheet, rows, row2, 1)
idx2 := -1
for i, r := range xlsx.SheetData.Row {
if r.R == row2 {
idx2 = i
break
}
}
if idx2 == -1 && len(xlsx.SheetData.Row) >= row2 {
return
}
rowCopy.C = append(make([]xlsxC, 0, len(rowCopy.C)), rowCopy.C...)
f.ajustSingleRowDimensions(&rowCopy, row2)
if idx2 != -1 {
xlsx.SheetData.Row[idx2] = rowCopy
} else {
xlsx.SheetData.Row = append(xlsx.SheetData.Row, rowCopy)
}
}
// checkRow provides a function to check and fill each column element for all
// rows and make that is continuous in a worksheet of XML. For example:
//
//
//
//
//
//
//
//
// in this case, we should to change it to
//
//
//
//
//
//
//
//
//
//
//
// Noteice: this method could be very slow for large spreadsheets (more than
// 3000 rows one sheet).
func checkRow(xlsx *xlsxWorksheet) {
for rowIdx := range xlsx.SheetData.Row {
rowData := &xlsx.SheetData.Row[rowIdx]
colCount := len(rowData.C)
if colCount == 0 {
continue
}
lastCol, _ := MustCellNameToCoordinates(rowData.C[colCount-1].R)
if colCount < lastCol {
oldList := rowData.C
newlist := make([]xlsxC, 0, lastCol)
rowData.C = xlsx.SheetData.Row[rowIdx].C[:0]
for colIdx := 0; colIdx < lastCol; colIdx++ {
newlist = append(newlist, xlsxC{R: MustCoordinatesToCellName(colIdx+1, rowIdx+1)})
}
rowData.C = newlist
for colIdx := range oldList {
colData := &oldList[colIdx]
colNum, _ := MustCellNameToCoordinates(colData.R)
xlsx.SheetData.Row[rowIdx].C[colNum-1] = *colData
}
}
}
}
// convertRowHeightToPixels provides a function to convert the height of a
// cell from user's units to pixels. If the height hasn't been set by the user
// we use the default value. If the row is hidden it has a value of zero.
func convertRowHeightToPixels(height float64) float64 {
var pixels float64
if height == 0 {
return pixels
}
pixels = math.Ceil(4.0 / 3.0 * height)
return pixels
}