// Package xlsx implements the sq driver for Microsoft Excel. package xlsx import ( "context" "database/sql" "io" "io/ioutil" "github.com/neilotoole/lg" "github.com/tealeg/xlsx/v2" "github.com/neilotoole/sq/libsq/core/cleanup" "github.com/neilotoole/sq/libsq/core/errz" "github.com/neilotoole/sq/libsq/core/options" "github.com/neilotoole/sq/libsq/driver" "github.com/neilotoole/sq/libsq/source" ) const ( // Type is the sq source driver type for XLSX. Type = source.Type("xlsx") ) // Provider implements driver.Provider. type Provider struct { Log lg.Log Files *source.Files Scratcher driver.ScratchDatabaseOpener } // DriverFor implements driver.Provider. func (p *Provider) DriverFor(typ source.Type) (driver.Driver, error) { if typ != Type { return nil, errz.Errorf("unsupported driver type %q", typ) } return &Driver{log: p.Log, scratcher: p.Scratcher, files: p.Files}, nil } var _ source.TypeDetectFunc = DetectXLSX // DetectXLSX implements source.TypeDetectFunc, returning // TypeXLSX and a score of 1.0 valid XLSX. func DetectXLSX(ctx context.Context, log lg.Log, openFn source.FileOpenFunc) (detected source.Type, score float32, err error) { var r io.ReadCloser r, err = openFn() if err != nil { return source.TypeNone, 0, errz.Err(err) } defer log.WarnIfCloseError(r) data, err := ioutil.ReadAll(r) if err != nil { return source.TypeNone, 0, errz.Err(err) } // We don't need to read all rows, one will do. const rowLimit = 1 _, err = xlsx.OpenBinaryWithRowLimit(data, rowLimit) if err != nil { return source.TypeNone, 0, nil } return Type, 1.0, nil } // Driver implements driver.Driver. type Driver struct { log lg.Log scratcher driver.ScratchDatabaseOpener files *source.Files } // DriverMetadata implements driver.Driver. func (d *Driver) DriverMetadata() driver.Metadata { return driver.Metadata{ Type: Type, Description: "Microsoft Excel XLSX", Doc: "https://en.wikipedia.org/wiki/Microsoft_Excel"} } // Open implements driver.Driver. func (d *Driver) Open(ctx context.Context, src *source.Source) (driver.Database, error) { r, err := d.files.Open(src) if err != nil { return nil, err } defer d.log.WarnIfCloseError(r) b, err := ioutil.ReadAll(r) if err != nil { return nil, errz.Err(err) } xlFile, err := xlsx.OpenBinary(b) if err != nil { return nil, err } scratchDB, err := d.scratcher.OpenScratch(ctx, src.Handle) if err != nil { return nil, err } clnup := cleanup.New() clnup.AddE(scratchDB.Close) err = xlsxToScratch(ctx, d.log, src, xlFile, scratchDB) if err != nil { d.log.WarnIfError(clnup.Run()) return nil, err } return &database{log: d.log, src: src, impl: scratchDB, files: d.files, clnup: clnup}, nil } // Truncate implements driver.Driver. func (d *Driver) Truncate(ctx context.Context, src *source.Source, tbl string, reset bool) (affected int64, err error) { // TODO: WE could actually implement Truncate for xlsx. // It would just mean deleting the rows from a sheet, and then // saving the sheet. return 0, errz.Errorf("source type %q (%s) doesn't support dropping tables", Type, src.Handle) } // ValidateSource implements driver.Driver. func (d *Driver) ValidateSource(src *source.Source) (*source.Source, error) { d.log.Debugf("Validating source: %q", src.RedactedLocation()) if src.Type != Type { return nil, errz.Errorf("expected source type %q but got %q", Type, src.Type) } return src, nil } // Ping implements driver.Driver. func (d *Driver) Ping(ctx context.Context, src *source.Source) (err error) { r, err := d.files.Open(src) if err != nil { return err } defer d.log.WarnIfCloseError(r) b, err := ioutil.ReadAll(r) if err != nil { return errz.Err(err) } _, err = xlsx.OpenBinaryWithRowLimit(b, 1) if err != nil { return errz.Err(err) } return nil } // database implements driver.Database. type database struct { log lg.Log src *source.Source files *source.Files impl driver.Database clnup *cleanup.Cleanup } // DB implements driver.Database. func (d *database) DB() *sql.DB { return d.impl.DB() } // SQLDriver implements driver.Database. func (d *database) SQLDriver() driver.SQLDriver { return d.impl.SQLDriver() } // Source implements driver.Database. func (d *database) Source() *source.Source { return d.src } // TableMetadata implements driver.Database. func (d *database) TableMetadata(ctx context.Context, tblName string) (*source.TableMetadata, error) { srcMeta, err := d.SourceMetadata(ctx) if err != nil { return nil, err } return source.TableFromSourceMetadata(srcMeta, tblName) } // SourceMetadata implements driver.Database. // TODO: the implementation of SourceMetadata is out // of sync with the way we import data. For example, empty // rows are filtered out during import, and empty columns // are discarded. Thus SourceMetadata needs an overhaul to // bring its reporting into line with import. func (d *database) SourceMetadata(ctx context.Context) (*source.Metadata, error) { meta := &source.Metadata{Handle: d.src.Handle} var err error meta.Size, err = d.files.Size(d.src) if err != nil { return nil, err } meta.Name, err = source.LocationFileName(d.src) if err != nil { return nil, err } meta.FQName = meta.Name meta.Location = d.src.Location meta.SourceType = Type b, err := d.files.ReadAll(d.src) if err != nil { return nil, errz.Err(err) } xlFile, err := xlsx.OpenBinary(b) if err != nil { return nil, errz.Errorf("unable to open XLSX file: ", d.src.Location, err) } hasHeader, _, err := options.HasHeader(d.src.Options) if err != nil { return nil, err } for _, sheet := range xlFile.Sheets { tbl := &source.TableMetadata{Name: sheet.Name, RowCount: int64(len(sheet.Rows))} if hasHeader && tbl.RowCount > 0 { tbl.RowCount-- } colNames := getColNames(sheet, hasHeader) // TODO: Should move over to using kind.Detector colTypes := getCellColumnTypes(sheet, hasHeader) for i, colType := range colTypes { col := &source.ColMetadata{} col.BaseType = cellTypeToString(colType) col.ColumnType = col.BaseType col.Position = int64(i) col.Name = colNames[i] tbl.Columns = append(tbl.Columns, col) } meta.Tables = append(meta.Tables, tbl) } return meta, nil } // Close implements driver.Database. func (d *database) Close() error { d.log.Debugf("Close database: %s", d.src) // No need to explicitly invoke c.impl.Close because // that's already added to c.clnup return d.clnup.Run() }