Turning on legacy OCR engine mode

[dmypstl]

Whitelist and blacklist are not implemented in version 4.0 (issue) and user patterns do not work (issue). In this issue people recommend turning on legacy OEM using --oem 0 option flag. This option is not part of configs, but rather belong to the engine itself, like language.

Could we please enable more ocr options as arguments in tesseract::tesseract(), including oem to be able to temporarily switch to older version of the engine.

[jeroen]

The main problem is that v3 uses a different format for training data than v4. So we would need to manage different sets of training data within a single r package.

[dmypstl]

I see. This is pretty unfortunate. Maybe you could tag latest stable version before engine update and invite people to optionally install it with remotes::install_github("ropensci/tesseract@v3.0.4") or whatever the last version was before it was updated to 4.0. I think found the relevant commit now, but it is pretty awkward to refer to it.

[jeroen]

The easiest way to install an old version of the R package is using MRAN snapshots:

install.packages('tesseract', repos = 'https://cran.microsoft.com/snapshot/2018-09-01/')

[jeroen]

The whitelist / blacklist options are now supported in tesseract 4.1 (on cran).