Instead of an association list and a type contract, let's use a hash
table and a structure. This is: more explicit, encapsulates state
more effectively, generates much of the cruft (constructors,
destructors) for me.
A hash table is more efficient when considering large player bases.
In particular, we'll be generating unique player IDs to make it a bit
more difficult for a malicious agent to guess another player's ID and
make bad requests.
Custom errors for malformed/nonexistent player IDs as well as a player
not having a high enough balance for the money requested.
This will allow me to encode more information in the errors, for
callers to use. I'm generally of the opinion that callers should be
checking error conditions before calling these functions, but per
Murphy's law it would be nice to have more information in the error
message.
This package defines a player data structure, collections of players,
and different interactions you can have with them:
- Extracting attributes (destructors)
- Checking if they're bankrupt or can bet
- Modifying a collection to adjust the balance of any one
player (including errors in case that isn't possible)
Joker constructor takes a rank for uniqueness - rank doesn't really
change anything in terms of the "power" of a joker but helps with
ensuring jokers are unique in a deck.
Deck constructor takes an optional argument for the number of decks.
A deck includes two jokers, so n decks include 2n jokers.
Given n in 0..51, certainly there exists r, s s.t. `n = 13s + r`
where r in 0..12 and s in 0..3. `r` is the rank, and `s` is the suit.
ranks are ordered Ace, 2, 3, ..., Queen, King and suits are ordered
Diamond, Club, Heart, Spade.
Therefore there is a 1-1 correspondence between 0..51 and any card in
the deck.
Jokers are a bit less straightforward; for now I'll define them as an
outlier which doesn't exist in the same bounds (which is why they're
the default return for suit in int->suit). Likely I'll go for -1
representing a Joker.
Instead of exporting cantedraw.lib.macro._ and making anyone who wants
to use the --> macro _require_ importing that specific symbol, let's
just make it so the user has to supply a placeholder name before they
do anything. This means each package provides its own placeholder
symbol which lowers coupling.
Given an indicator function (A->B) and a list of items of A, return
an association list associating B to the elements that map to it;
essentially the inverse map of the indicator.
The `$` operator takes a sequence of FORMS and returns a unary
function which applies the input through that sequence via the `->>`
operator.
For example, consider the predicate "not null". `null` is built into
Common Lisp but "not null" requires writing a
function (lambda (x) (not (null x))). Now, using this operator, you
can write ($ not null) which returns the same lambda as above while
being more concise.
Splitting macros and functions into different packages and source code
makes it easier to look at. Functions currently implemented:
- range: like Python's range
- parse-integer*: parse-integer but junk-allowed is set to t.
`fn' is a convenience macro for defining functions with a type
specifier. Only really matters for `sbcl` and other hard-optimising
Lisp interpreters which actually take these seriously.
Setup boilerplate for system/package management. In particular, setup
an entry-point and Shinmera's "deploy" to build executables.
Also write some scripts to easily load or build the project without
Emacs - just `sbcl --load <x>.lisp`.
