Sobre os projetos, sugero que os alunos invitam em projetos que possam ser disponibilizados como bibliotecas Racket
Isto seria um excelente resultado para os projetos além de apenas um relatório acompanhado de um código que não terá uso depois do curso. Alternativamente, os grupos podem tentar contribuir com algum projeto já existente. Notem que para efetivamente distribuir um código que potencialmente poderá ser organizado em vários arquivos, será necessário entender o que são módulos e pacotes:
https://docs.racket-lang.org/guide/modules.html
Como eu disse em sala, o LoL (https://goo.gl/H2qnNs) usa values para
isso. Para saber mais como lidar com values, vejam
https://goo.gl/IitGiv.
(let ((counter 0))
(values
(lambda () (incf counter))
(lambda () (decf counter))))E values também existe em Racket.
> (values 1 2) 1 2
Sobre https://goo.gl/H2qnNs, vale muito ler o trecho abaixo, ver a solução do block-scanner e pensar sobre:
Unfortunately, most lisp programming books don’t provide realistic examples of closure usage, leaving readers with the inaccurate impression that closures are only good for toy examples like counters. Nothing could be further from the truth. Closures are the building blocks of lisp. Environments, the functions defined inside those environments, and macros like defun that make using them convenient, are all that are needed for modelling any problem. This book aims to stop beginning lisp programmers used to object-based languages from acting upon their gut instinct of reaching for systems like CLOS. While CLOS does have certain things to offer the professional lisp programmer, do not use it when a lambda will suffice.
In order to motivate the use of closures, a realistic example is presented: block-scanner. The problem block-scanner solves is that for some forms of data transfer the data is delivered in groups (blocks) of uncertain sizes. These sizes are generally convenient for the underlying system but not for the application programmer, often being determined by things like operating system buffers, hard drive blocks, or network packets. Scanning a stream of data for a specific sequence requires more than just scanning each block as it comes in with a regular, stateless procedure. We need to keep state between the scanning of each block because it is possible that the sequence we are scanning for will be split between two (or more) blocks.