[racket] Use of map and eval to evaluate symbol in namespace

On Aug 3, 2014, at 4:29 PM, Alexander D. Knauth <alexander at knauth.org> wrote:

> 
> On Aug 3, 2014, at 3:29 PM, Henry Lenzi <henry.lenzi at gmail.com> wrote:
> 
>> ; Hello all --
>> ; So here's how I solve all those little problems regarding symbols
>> and evaluation of medication definitions.
>> ; Would you please bear with me? I apologize for the length.
>> ; This is the approach I've taken. I've chosen no to use any macrology
>> or parser/lexer technique because I don't grok them and they
>> ; don't really seem necessary, for reasons explained in the code comments.
>> ; I have not decided to hash tables, for the following reason: there's
>> a part of the code (the drug definitions, the instructions), that
>> ; should be easy enough for non-programmers to edit. If they are kept
>> very simple, it's possible, because the users have to edit those
>> ; files. So, even though it is source code, it's not as intimidating
>> as editing source code if hash tables.
>> ; Another aspect is that I hope modules provided some sort of safety
>> in terms of syntax checking. That is to say, if you used make a
>> ; typo in the medication part of the DSL, the system will (hopefully)
>> bork because no such module exists. I believe this also creates
>> ; an opportunity for "syntax validation" if a proper input phase is
>> designed. But Lisp/Scheme being a dynamic language, the run-time
>> ; will bork immediately once it sees funny things. This is a way to
>> guarantee the DSL is correct, which we get for free by using Racket.
>> ; A fourth aspect is that, if each drug is kept a different module
>> (which I haven't done here, BTW), then we can make for easier
>> ; internationalization, by keeping modules by languages, e.g.,
>> hctz25-en, hctz25-pt_br. I believe Dan has an interest in this project
>> too,
>> ; so  it's best to design with that in mind.
>> ; Final comment regards "database". We get "database" for free, by
>> registering prescriptions with patient register numbers. The OS
>> ; takes care of pretty musch anything else. And there's no need for
>> atomicity and concurrency. Like I said, this is stupid code.
>> ;
>> ;
>> #lang racket
>> 
>> ; code-review-for-racketeers-2014-08-03-a.rkt
>> ;
>> ; For this exercise, suppose a Recipe.txt file. Let´s suppose the idea
>> is that the physician
>> ; has two options: 1) he or she opens Notepad and writes the
>> prescription file (Recipe.text);
>> ; 2) or, the software asks for inputs and writes the file (this will
>> not be covered in this
>> ; exercise). The written prescription in the shorthand DSL would look
>> like below, with the
>> ; exception of a first field with patient ID data not included (to be
>> done later).
>> ; The prescription has a rigid syntax would look like this (line
>> breaks included):
>> ; 1-
>> ; hctz25 30 pl 1xd
>> ;
>> ; 2-
>> ; simva20 30 pl 1xn
>> 
>> 
>> ; Needed for EVAL, used later on
>> (define-namespace-anchor a)
>> 
>> ; These definitions should be in a different module.
>> ; This way we get syntax checking for free.
>> ; MED - medication. Includes dosage.
>> (define hctz25 "Hydrochlorothiazide 25mg")
>> (define simva20 "Simvastatin 20mg")
>> ; FORM - whether the patient will take home pills, a tube, a flask, capsules
>> (define pl "pills")
>> ; POS - posology, whether the patient will take 1 pill 3x a day, or 2
>> pills 2x a day, etc.
>> (define 1xd "Take 1 pill P.O. 1x/day")
>> (define 1xn "Take 1 pill P.O. 1x at night")
>> ; INSTs - special instructions. INST is just a prefix INST+MED without
>> the dosage.
>> (define INSTOMZ "half an hour before breakfast, with a glass of water")
>> ; Formatters - simple for now, but should be a function of the space available.
>> (define  line "-----------")
>> 
>> 
>> 
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>> ; The main part of a prescription DSL is pretty rigid in syntax, being
>> composed of blocks of theses parts:
>> ; MEDICATION QUANTITY FORM POSOLOGY INSTRUCTION, or MED QUANT FORM POS INST.
>> ; Please note that, in this DSL, the MED part includes the drug dosage
>> (e.g., HCTZ25, where
>> ; the HCTZ designates the drug, and the 25 the dosage).
>> ; An example would be:
>> ; HCTZ25 30 PL 1XD
>> ; meaning: Hydrochlorothiazide 25mg -------------- 30 pills
>> ;                Take 1 pill P.O. 1X day
>> ; INST are special instructions. They basically are more detailed
>> explanation to the patient about
>> ; how to use the medication properly. Not always there's a INST in the
>> prescription DSL.
>> ; INSTs are, in fact, a PREFIX for the MED without the dose. For
>> example, OMZ20 is Omeprazol 20mg.
>> ; The instruction for OMZ would be INSTOMZ ("half an hour before
>> breakfast, with a glass of water").
>> ; In this case, the DSL line would be:
>> ; OMZ20 30 PL 1XD INSTOMZ
>> ; meaning: Omeprazol 20mg ------------------- 30 pills
>> ;               Take 1 pill P.O. 1X day
>> ;               half an hour before breakfast, with
>> ;               a glass of water
>> ; Questions regarding proper formatting of INST are not addressed at
>> this moment.
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>> ; Now follows a description of some problems I encountered and the
>> choices made in solving them:
>> ; (define in (open-input-file "Recipe.txt"))
>> ; If you just (string-split (read-line in)) you'll get:
>> ; => '("hctz25" "30" "cp" "1xd")
>> ; and that will not evaluate the symbols to their string descritptions.
>> ; Because of that, you need to do a:
>> ; > (map string->symbol (string-split (read-line in)))
>> ; which will evaluate to
>> ; => '(hctz25 |30| cp 1xd)
>> ; This would be ideal to MAP EVAL to, but the problem is the |30|
> 
> What you want here is something like this:
> ;; Instead of (map string->symbol (string-split (read-line in)))
> (for/list ([thing (in-port read in)]) thing)

Actually come to think of it you can do this:
(sequence->list in)
Or this:
(port->list read in)

> ;; and then you can do (map eval …) to that if you want.
> ;; Or you could do both at once like this:
> (for/list ([thing (in-port read in)])
>  (eval thing namespace))

Or for that:
(port->list (compose1 eval read) in)

> 
>> ; So, the idea is SET!ing that list to a name we can call easily, i.e.,
>> ; med-line-holder, because then we can extract the pieces (since we
>> can't do list
>> ; surgery easily, such a "replace the the element at position 1 with
>> so-and-so element”).

look at list-set from unstable/list

>> ; Since the prescription syntax is pretty rigid, we can get away with this
>> ; simple approach.
>> 
>> (define med-line-holder '()) ; initial value of med-line-holder is an empty list
>> (define med-name-holder '())
>> (define med-quant-holder '())
>> (define med-form-holder '())
>> (define med-pos-holder '())
>> (define med-inst-holder '()) ; remember, not always INSTructions
>> happen in a DSL prescription .
>> 
>> (define in (open-input-file "Recipe.txt"))
>> (port-count-lines! in)
>> (define (clpr) (close-input-port in))
>> 
>> ; a med-line-holder is a list that has MED QUANT FORM POS (and sometimes INST)
>> ; This is obtained from a plain text file. When it is read, it becomes something
>> ; like this: '(hctz25 |30| cp 1xd)
>> (define (set-med-line-holder)
>> (set! med-line-holder (map string->symbol (string-split (read-line in)))))
>> 
>> (define (set-med-name-holder)
>>  ; (set! med-name-holder (eval (car med-line-holder))) ;; in the REPL
>> (set! med-name-holder (eval (car med-line-holder)
>> (namespace-anchor->namespace a))))
>> 
>> (define (set-med-quant-holder) ; the CADR of the med-line-holder
>> ; (set! med-quant-holder (eval (symbol->string (cadr med-line-holder))))
>> (set! med-quant-holder (eval (symbol->string (cadr med-line-holder))
>> (namespace-anchor->namespace a))))
>> 
>> (define (set-med-form-holder) ; the CADDR of the med-line-holder -
>> gets the FORM, e.g., pills, etc.
>> ; (set! med-form-holder (eval (symbol->string (caddr med-line-holder))))
>> (set! med-form-holder (eval (caddr med-line-holder)
>> (namespace-anchor->namespace a))))
>> 
>> (define (set-med-pos-holder) ; the CADDDR of the med-line-holder -
>> gets the POS, e.g., 1xd
>>   ; (set! med-pos-holder (eval (symbol->string (cadddr med-line-holder))))
>> (set! med-pos-holder (eval (cadddr med-line-holder)
>> (namespace-anchor->namespace a))))
>> 
>> 
>> (define (set-med-inst-holder) ; the LAST of the med-line-holder - gets the INST
>>   ; (set! med-pos-holder (eval (symbol->string (last med-line-holder))))
>> (set! med-pos-holder (eval (last med-line-holder)
>> (namespace-anchor->namespace a))))
>> 
>> ; One problem here regards the optional INST instructions.
>> ; How to create a SETter function that will only SET! med-inst-holder
>> ; if there's an INST instruction? Note that INST is a prefix. A real
>> instruction is, e.g.,
>> ; INSTOMZ (for OMZ20).
>> (define (look-for-line)
>> (if (regexp-match #px"\\d\\-" (read-line in))
>>     (begin
>>       (set-med-line-holder)
>>       (set-med-name-holder)
>>       (set-med-quant-holder)
>>       (set-med-form-holder)
>>       (set-med-pos-holder))
>>     'NO-LINE))
>> 
>> (define (display-stuff)
>> (newline)
>> (display med-line-holder) (newline)
>> (display med-name-holder) (newline)
>> (display med-quant-holder) (newline)
>> (display med-form-holder) (newline)
>> (display med-pos-holder) (newline))
>> ; The problem remains of what to do with the eventual INST.
>> 
>> 
>> ; Successive calls to (look-for-line) would read the next lines.
>> ; Output would alternate between a DSL line, or a NO-LINE (from look-for-line,
>> ; if it hits a line with no text in Recipe.txt
>> (look-for-line)
>> ;(display-stuff)
>> 
>> 
>> (define (output-a-line)
>> (string-join (list med-name-holder line med-quant-holder med-form-holder "\n"
>>                            med-pos-holder "\n")))
>> 
>> (define (format-a-line)
>> (display (output-a-line)))
>> 
>> ;(define (output-a-line)
>> ; (display (string-join (list med-name-holder line med-quant-holder
>> med-form-holder "\n"
>> ;                             med-pos-holder "\n"))))
>> (newline)
>> ;(output-a-line)
>> 
>> (format-a-line)
>> 
>> 
>> 
>> ; PROBLEMS
>> ; 1) How do we find out how many lines to (look-for-line)?
>> ;    This is one of the resons I specified the "1-", "2-" in the Recipe.txt. Not
>> ;    only it makes for easy visual understanding, but it may be used
>> to provide a hint
>> ;    for this problem.
>> ;    Possible approaches:
>> ;    - Maybe this can be solved with REGEXPS? This information could
>> provide a sentinel
>> ;      variable for an iterator function?
>> ;    - Is there some sort if line counting function? (Note that I have set
>> ;      (port-count-lines! in) somewhere above in the code.
>> ; 2) How do we know we've reached the end of the file?
>> ; 3) How to deal with the not-always-present INST?
>> ;    - How do we check for INSTs? With a REGEXP?
>> ;    - Choosing between INSTs with REGEXPS is not necessary, as they
>> will be loaded in a module,
>> ;      so the system will "know" which one to choose.
>> ; 4) Another idea would be "slurp" the whole of the prescription, and
>> then deal with evaluation. How?
>> ; (define f1
>> ;    (file->string
>> ;   "C:\\Path\\to\\sources\\Recipe.txt"))
>> ;> (string-normalize-spaces f1)
>> ;"1- hctz25 30 pl 1xd 2- simva20 30 pl 1xn"
>> ;
>> ; That's all for now, folks!
>> ; Many thanks for all the help so far, Racketeers!
>> ; Cheers,
>> ; Henry Lenzi
>> 
>> 
>> 
>> 
>> On Sat, Aug 2, 2014 at 12:44 AM, Henry Lenzi <henry.lenzi at gmail.com> wrote:
>>> Hello everyone -
>>> 
>>> First of all, a big Thank You to all of you and for taking the time for
>>> responding.
>>> 
>>> I'll have to set aside sometime during this weekend to see if I can
>>> understand the ideas you've been so kind to offer.
>>> 
>>> However, I should confess that I've made some progress with way simpler
>>> stuff which I hope to post later on.  Like I've said, this is stupid
>>> software. Anyways, none of this is final.
>>> 
>>> It really just used a plain text solution, since the format if a recipe is
>>> so rigid. The question of expanding the symbols from files to run-time was
>>> easier than I thought.
>>> 
>>> The idea of using modules might have the nice collateral effect if some sort
>>> of primitive type (or syntax) checking for free. I like the idea someone
>>> offered of using modules for medication definitions. Actually, one module
>>> per definition makes it very easy for future users to add new medications.
>>> The ease of syntax is important because it allows for the customization by
>>> non-sophisticated users (physicians, nurses).
>>> 
>>> Cheers,
>>> Henry Lenzi.
>> 
>> ____________________
>> Racket Users list:
>> http://lists.racket-lang.org/users
> 
> 
> ____________________
>  Racket Users list:
>  http://lists.racket-lang.org/users