money/internal/parser/cmd.go

package parser

import (
	"bytes"
	"context"
	"encoding/csv"
	"fmt"
	"io"
	"os/exec"
	"strings"
	"time"

	"git.petrovv.com/nikola/money/internal/config"
)

func init() {
	Register("cmd", newCmdParser)
}

// FileToken is replaced with the statement's absolute path in a cmd parser's argv.
const FileToken = "{{file}}"

// cmdRunTimeout bounds an extractor run so a hung script cannot wedge an import.
const cmdRunTimeout = 2 * time.Minute

// cmdParser runs an external extractor and reads normalised CSV from its
// stdout: date,description,amount with any further columns ignored. This is
// the bridge that lets the existing Python extractors be used unchanged.
type cmdParser struct {
	cfg    config.CmdConfig
	digits int
}

func newCmdParser(acc *config.Account) (Parser, error) {
	if acc.Cmd == nil || len(acc.Cmd.Argv) == 0 {
		return nil, fmt.Errorf("account %s: parser \"cmd\" requires [cmd] with a non-empty argv", acc.Slug)
	}
	cfg := *acc.Cmd
	if cfg.Layout == "" {
		cfg.Layout = "2006-01-02"
	}
	if !hasFileToken(cfg.Argv) {
		return nil, fmt.Errorf("account %s: [cmd] argv must contain %s so the script knows which file to read",
			acc.Slug, FileToken)
	}
	return &cmdParser{cfg: cfg, digits: acc.Digits()}, nil
}

func hasFileToken(argv []string) bool {
	for _, a := range argv {
		if strings.Contains(a, FileToken) {
			return true
		}
	}
	return false
}

func (p *cmdParser) Parse(path string, acc *config.Account) ([]RawTxn, error) {
	argv := make([]string, len(p.cfg.Argv))
	for i, a := range p.cfg.Argv {
		argv[i] = strings.ReplaceAll(a, FileToken, path)
	}

	ctx, cancel := context.WithTimeout(context.Background(), cmdRunTimeout)
	defer cancel()

	cmd := exec.CommandContext(ctx, argv[0], argv[1:]...)
	cmd.Dir = acc.Dir
	var stdout, stderr bytes.Buffer
	cmd.Stdout = &stdout
	cmd.Stderr = &stderr

	if err := cmd.Run(); err != nil {
		msg := strings.TrimSpace(stderr.String())
		if ctx.Err() == context.DeadlineExceeded {
			return nil, fmt.Errorf("extractor %v timed out after %s", argv, cmdRunTimeout)
		}
		if msg != "" {
			return nil, fmt.Errorf("extractor %v failed: %w: %s", argv, err, msg)
		}
		return nil, fmt.Errorf("extractor %v failed: %w", argv, err)
	}

	return p.parseOutput(&stdout, argv)
}

func (p *cmdParser) parseOutput(out io.Reader, argv []string) ([]RawTxn, error) {
	r := csv.NewReader(out)
	r.FieldsPerRecord = -1
	r.LazyQuotes = true

	var txns []RawTxn
	for row := 0; ; row++ {
		rec, err := r.Read()
		if err == io.EOF {
			break
		}
		if err != nil {
			return nil, fmt.Errorf("extractor %v: output row %d: %w", argv, row+1, err)
		}
		if row < p.cfg.SkipRows || isBlank(rec) {
			continue
		}
		if len(rec) < 3 {
			return nil, fmt.Errorf("extractor %v: output row %d has %d columns, want date,description,amount",
				argv, row+1, len(rec))
		}
		d, err := time.Parse(p.cfg.Layout, strings.TrimSpace(rec[0]))
		if err != nil {
			return nil, fmt.Errorf("extractor %v: output row %d: date %q does not match layout %q",
				argv, row+1, rec[0], p.cfg.Layout)
		}
		amount, err := ParseAmount(rec[2], ".", "", p.digits)
		if err != nil {
			return nil, fmt.Errorf("extractor %v: output row %d: %w", argv, row+1, err)
		}
		txns = append(txns, RawTxn{
			Date:        d.Format("2006-01-02"),
			Description: strings.TrimSpace(rec[1]),
			AmountMinor: amount,
		})
	}
	return txns, nil
}