[wleeaf]

pdf417.Encode puts short data at a low security level into a symbol with only 2 rows, for example Encode("A", 0), Encode("A", 1) and Encode("Hell", 0). ISO/IEC 15438 allows 3 to 90 rows, and readers don't find 2-row symbols: zxing-cpp, ZXing (Java) and rxing all fail on them.

This raises minRows to 3. A message that would fit in fewer rows now gets the 3-row, 2-column fallback size, so getPadding pads to columns * rows codewords instead of to the next multiple of columns. For every size the search loop picks, the two amounts are the same, so other symbols don't change.

TestMinimumRows encodes short strings at security levels 0 to 2 and checks the row count. On master it fails for 7 of them, which come out with 2 rows. go test ./..., go vet ./... and golangci-lint run pass.

I also decoded 910 PDF417 symbols with zxing-cpp: digits, text and alphanumerics of lengths 1 to 300 at security levels 0, 2 and 5, plus a few special inputs. On master 6 of them are not found. With this change all of them decode to the input, and ZXing (Java) and rxing read all 910 as well.

[danielwhite]

@wleeaf, thanks for the report. This was on my radar, but was scheduled behind further regression testing to land this week in preparation for a v1.2.0 release.

I'll follow up once that's complete and take a closer look.