Hi all, I'm working on a Haskell (Eta) tool to generate Java ADTs. I have the following definitions: data Line = Line Int Text [JavaImport] -- The Int represents the indentation level of this line. data Template = None | One Line | Many [Line] | Template (Reader Configuration Template) I then run it like this (incomplete, I hope you get the idea): main :: IO () main = putStr $ replace (importsGoHereKey `append` "\n") importsAsText templateAsText where config = Configuration (Copyright 2018 "Me") USE_GETTERS point' = javaClassTemplate point importsAsText = runReader (mapReader javaImportsToText (javaImports point')) config templateAsText = runReader (toText tab point') config Essentially, I use the template twice: once to extract the Java imports and once to generate the result Text. Then I replace a magic string in the result Text with the import code. It works but makes me sad. ;-) I will probably pull out the JavaImport list from Line and use something like State to keep track of the necessary imports. (Now that I understand Reader, maybe State is no longer a bridge too far.) My main issue is how to avoid the magic string. I thought about adding an extra entry to Template [something like Imports ((Set JavaImport) -> Text or maybe Template)] but then I don't know how to implement "instance Monoid Template". Similarly if I use something Reader like. Any design advice or ideas would be appreciated. If it's of any help, the full code is at [1] (it's tiny). Any and all comments are welcome, by the way: I'm still a Haskell beginner. Cheers, Hilco [1] https://github.com/hilcode/java-algebraic-data-types