about summary refs log tree commit diff
path: root/web/panettone/src/inline-markdown.lisp
diff options
context:
space:
mode:
Diffstat (limited to 'web/panettone/src/inline-markdown.lisp')
-rw-r--r--web/panettone/src/inline-markdown.lisp130
1 files changed, 130 insertions, 0 deletions
diff --git a/web/panettone/src/inline-markdown.lisp b/web/panettone/src/inline-markdown.lisp
new file mode 100644
index 000000000000..fc5f77584fc4
--- /dev/null
+++ b/web/panettone/src/inline-markdown.lisp
@@ -0,0 +1,130 @@
+(in-package :panettone.inline-markdown)
+(declaim (optimize (safety 3)))
+
+(define-constant +inline-markup-types+
+  '(("~~" :del)
+    ("*"  :em)
+    ("`"  :code))
+  :test #'equal)
+
+(defun next-token (mkdn &optional (escaped nil))
+  "Parses and returns the next token from the beginning of
+  an inline markdown string which is not altered. The resulting
+  tokens are either :normal (normal text), :special (syntactically
+  significant) or :escaped (escaped using \\). If the string is
+  empty, a pseudo-token named :endofinput is returned. Return value
+  is a list where the first element is the token type, the second
+  the token content and optionally the third the markup type."
+  ; special tokens are syntactically significant characters
+  ; or strings for our inline markdown subset. “normal” tokens
+  ; the strings in between
+  (let* ((special-toks #.'(cons (list "\\" :escape) +inline-markup-types+))
+         (toks (loop
+                 for tok in special-toks
+                 for pos = (search (car tok) mkdn)
+                 when pos collect (cons tok pos)))
+         (next-tok
+           (unless (null toks)
+             (reduce (lambda (a b) (if (< (cdr a) (cdr b)) a b)) toks))))
+    (cond
+      ; end of input
+      ((= (length mkdn) 0) (list :endofinput ""))
+      ; no special tokens, just return entire string
+      ((null next-tok) (list :normal mkdn))
+      ; special token, but not at the beginning of the string
+      ; so we return everything until the special token as
+      ; a string
+      ((> (cdr next-tok) 0) (list :normal (subseq mkdn 0 (cdr next-tok))))
+      ; \ at the beginning of the string: we get the next
+      ; token and mark it as escaped unless we are already
+      ; escaping in which case we just return the backslash
+      ; as a special token
+      ((eq (cadr (car next-tok)) :escape)
+       (if escaped
+         (list :special "\\")
+         (list :escaped
+               (next-token (subseq mkdn 1) t))))
+      ; any other special token at the beginning of the string
+      ; here we also pass the markup type as a third list element
+      ; to prevent unnecessesary lookups
+      (t (list :special
+               (subseq mkdn 0 (length (car (car next-tok))))
+               (cadr (car next-tok)))))))
+
+(defun token-length (tok-type tok-str)
+  "Returns the string length consumed by a call
+  to next-token returning the given token type and string."
+  (check-type tok-type symbol)
+  (if (eq tok-type :escaped)
+    ; backslash + length of escaped token
+    (progn
+      (check-type tok-str list)
+      (1+ (token-length (car tok-str) (cadr tok-str))))
+    (progn
+      (check-type tok-str string)
+      (length tok-str))))
+
+(defun write-tag (tag pos &optional (target *standard-output*))
+  "Wrapper around who:convert-tag-to-string-list to
+  only output a single :opening or :closing tag."
+  (check-type tag symbol)
+  (check-type pos symbol)
+  (let
+    ((index
+       (cond
+         ((eq pos :opening) 0)
+         ((eq pos :closing) 3)
+         (t (error 'simple-type-error)))))
+    (dolist
+      (tag-part (subseq
+                  (who:convert-tag-to-string-list tag nil nil nil)
+                  index (+ index 3)))
+      (write-string tag-part target))))
+
+(defun render-inline-markdown (s &optional (target *standard-output*) (in :normal))
+  "Render inline markdown, a subset of markdown safe to render
+  inside inline elements. The resulting html is directly written
+  to a specified stream or *standard-output* to integrate well
+  with cl-who."
+  (check-type s string)
+  (check-type target stream)
+  (loop
+    for (tok-type tok-str tok-markup) = (next-token s)
+    do (setq s (subseq s (token-length tok-type tok-str)))
+    when (eq tok-type :endofinput)
+    return ""
+    when (eq tok-type :normal)
+    do (write-string (who:escape-string tok-str) target)
+    when (eq tok-type :escaped)
+    do (progn
+         ; if normal tokens are escaped we treat the \ as if it were \\
+         ;
+         ; TODO(sterni): maybe also use the :normal behavior in :code except for #\`.
+         (when (eq (car tok-str) :normal)
+           (write-char #\\ target))
+         (write-string (who:escape-string (cadr tok-str)) target))
+    when (eq tok-type :special)
+    do (cond
+         ; we are on the outer level and encounter a special token:
+         ; render surrounding tags and call ourselves to render
+         ; inner content.
+         ((eq in :normal)
+          (progn
+            (write-tag tok-markup :opening target)
+            (setq s (render-inline-markdown s target tok-markup))
+            (write-tag tok-markup :closing target)))
+         ; we are on the inner level and encounter the token that initiated
+         ; our markup again, meaning we need to return to the outer level.
+         ; we return the remaining string to be consumed.
+         ((eq in tok-markup) (return s))
+         ; remaining case: we are on the inner level and encounter different markup.
+
+         ; we don't support nested markup for simplicity reasons, so instead we
+         ; just render any nested markdown tokens as if they were escaped. This
+         ; only eliminates the slight use case for nesting :em inside :del, but
+         ; shouldn't be too bad. As a side effect this is the precise behavior
+         ; we want for :code.
+         ;
+         ; TODO(sterni): maybe bring back the restart-based system which allowed
+         ;               to skip nested tokens if desired.
+         (t (write-string (who:escape-string tok-str) target)))))