Performance Guide

Here we introduce some performance optimization tips for several scenarios.

Calling Multiple Range Methods

When calling Range setter method, Keikai spreadsheet will automatically check cell dependencies, update the dependent cells and refresh the spreadsheet UI of the range. However, in following cases, developers might not want such “automation” and like to control the evaluation and update by themselves:

  • Change a lot of cells in a batch.
  • If we don’t disable auto refresh in such case, the Keikai spreadsheet will generate a lot of small AU response to a browser which slows down browser rendering speed.
  • initialize a book upon a data source (e.g. a database) before Keikai spreadsheet renders itself.
  • Sometimes we need to load the data from a database to initialize a sheet before Keikai spreadsheet renders in a browser. Disable the auto refresh can eliminate Spreadsheet’s unnecessary internal calculations for rendering.

In order to manually control UI update, we have to:

  1. disable auto-refresh with setAutoRefresh(false)
  2. notify changed area with notifyChange()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
    private void loadData() {
        Sheet sheet = ss.getSelectedSheet();
        for (int column  = 0 ; column < COLUMN_SIZE ; column++){
            for (int row = 0 ; row < ROW_SIZE ; row++ ){
                Range range = Ranges.range(sheet, row, column);
                range.setAutoRefresh(false);
                range.getCellData().setEditText(row+", "+column);
                CellOperationUtil.applyFontColor(range, "#0099FF");
                CellOperationUtil.applyAlignment(range, Alignment.CENTER);
            }
        }
        Ranges.range(ss.getSelectedSheet(), 0, 0, ROW_SIZE, COLUMN_SIZE).notifyChange();
                ...
    }
  • line 6: disable the auto-refresh before changing cells (calling setter)
  • line 12: notify the changed range of cells or just the whole sheet

You can run Example Source to know the speed difference between 2 cases.

Notify Affected Range

When notifying a change, remember to choose an affected range, not just those cells you modify. The following cases explain the reasons:

  • Change a cell referenced by a formula in another cell.
  • If you change a cell, all those cells that contain a formula referencing the cell should also require an update.
  • Insertion / deletion of cells / rows / columns.
  • If you insert a column, all columns after the inserted column also require an update.

Notify the whole sheet

If the affected cells are too distributed, you can consider notifying the whole sheet. But this might make a sheet blank for a moment because it will re-render the whole sheet.

1
Ranges.range(ss.getSelectedSheet()).notifyChange();

Notify the cached area

If rendering a whole sheet is too slow, you can also consider to notify the currently cached area.

1
2
3
Spreadsheet ss;
// change cells
ss.notifyLoadedAreaChange();

No Auto-Adjusting Row Height

If your application doesn’t allow users to do any operation that needs a row height calculation e.g. enable / disable wrap text, change a font size, then you can set the attribute ignoreAutoHeight to true. This will improve client-side rendering speed much because it avoids time-consuming cell’s height calculating for each row.

1
2
<!-- default is false -->
<spreadsheet  ignoreAutoHeight="true"/>

Initialize with Large Data

Implement PostImport

A typical use case is loading a template file and inserting application data from a database at the beginning. Normally, this will generate lots of internal events and trigger formula dependency recalculation which is unnecessary before showing Keikai spreadsheet to a browser. You can implement

io.keikai.api.PostImport

and put your initialization logic in process(). Then Keikai will invoke process() right after the file importing and turn off those unnecessary update triggered by Range API. Therefore, it can speed up the data/formula inserting.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
public class PostImportComposer extends SelectorComposer<Component> implements PostImport{

    @Wire
    private Spreadsheet ss;
    @Wire("checkbox")
    private Checkbox postImportingBox;
    private String src = "/WEB-INF/books/blank.xlsx";
    private final File FILE = new File(WebApps.getCurrent().getRealPath(src));
    private Importer importer = Importers.getImporter();
    private String POST_IMPORT_KEY = "post-import";

    @Override
    public void doAfterCompose(Component comp) throws Exception {
        super.doAfterCompose(comp);
        long start = System.currentTimeMillis();
        if (isPostImported()){
            loadWithPostImporting();
        }else{
            loadDirectly();
        }
        long end = System.currentTimeMillis();
        postImportingBox.setChecked(isPostImported());
        Clients.showNotification("consumed (ms):"+(end-start));
    }

    @Override
    public void process(Book book) {
        initializeMassiveFormulas(book.getSheetAt(0));
    }

    /**
     * Increase row and column here, you will see bigger time difference between post importing and non-post importing. 
     * @param sheet
     */
    private void initializeMassiveFormulas(Sheet sheet){
        for (int row = 0 ; row < 1500 ; row++){
            for (int col = 0 ; col < 50 ; col++){
                String editText; 
                if (row > 0 ){
                    editText = "=" +Ranges.getCellRefString(row-1, col);
                }
                else{
                    editText = ""+(row+col);
                }
                Ranges.range(sheet, row, col).setCellEditText(editText);
            }
        }
    }
    
    private void loadWithPostImporting() throws IOException{
        Book book = importer.imports(FILE, "blank", this);
        ss.setBook(book);
    }
    
    private void loadDirectly() throws IOException{
        Book book = importer.imports(FILE, "blank");
        ss.setBook(book);
        initializeMassiveFormulas(ss.getSelectedSheet());
    }
...
}

Initialize Asynchronously

If the data to insert is too large, so that it still consumes a long time you can’t accept. Then you can insert the data in 2 phases. First, just insert a small part of data (e.g. 500 rows) since Keikai doesn’t render all rows to a browser and a user’s screen size is also limited. A user can’t see all rows at once in the beginning. Then send an Echo Event to insert the rest of the data.

Import a Large File

The major factor that determines the importing time is the number of cell instead of the number of sheet. Importing One file haveing 1 sheet, 10 thounsands cells takes longer time than another file having 20 sheets with 1 thousand cells in total.

Clear unnecessary cells

If one sheet actually contains 100 rows of data. But you apply cell background color to 5000 rows, then keikai still needs to process those (unneecessary) 4900 rows which is a waste of time. Hence, you can reduce the importing time by:

  • clear/delete those unnecessary cells
  • move those cells with data into a new sheet

Split the File

If you have a big file with multiple sheets and massive cells, and it takes a long time to import.You can split the file into multiple smaller files and import them separately. For example, you split a big file into a-1.xlsx and a-2.xlsx. Then, import the 1st file first, after keikai shows the first file, starts to import the 2nd file. Keep the reference of the 2 Book. Allow users to switch among 2 books by calling Spreadsheet::setBook().

Comments