summaryrefslogtreecommitdiff
path: root/tnslc/parse
diff options
context:
space:
mode:
authorKyle Gunger <kgunger12@gmail.com>2024-10-02 16:39:47 -0400
committerKyle Gunger <kgunger12@gmail.com>2024-10-02 16:41:16 -0400
commit442986c2c48b4653388b30b74e8c1ccd11423578 (patch)
treee2103568f40640177d3d0b4b72d9d1e0d1c4d52e /tnslc/parse
parent28f480a378e8c62cb7dbc649116d3505edf4fe7b (diff)
[tnslc] First draft of _mhf_post_list. Probably BUGGY!
Diffstat (limited to 'tnslc/parse')
-rw-r--r--tnslc/parse/ast.tnsl131
1 files changed, 126 insertions, 5 deletions
diff --git a/tnslc/parse/ast.tnsl b/tnslc/parse/ast.tnsl
index 7ce321b..7f10b6a 100644
--- a/tnslc/parse/ast.tnsl
+++ b/tnslc/parse/ast.tnsl
@@ -71,9 +71,18 @@ struct Node {
/; add_child(~Node n) [~Node]
n`.parent = ~self
+ ~int chk = self.sub.data
self.sub.push(n)
- /; loop (int i = 0; i < self.sub.count) [i++]
- n = self.sub.get(i)
+ ~int cmp = self.sub.data
+ /; if (chk != cmp)
+ # Only update children when the reallocation
+ # returns a new memory block
+ /; loop (int i = 0; i < self.sub.count) [i++]
+ n = self.sub.get(i)
+ n`.update_children()
+ ;/
+ ;; else
+ n = self.sub.get(self.sub.count - 1)
n`.update_children()
;/
return n
@@ -734,7 +743,7 @@ struct Node {
bool seen = false
/; loop (first`._type !== TTYPE_ERR && first`.data` !== end)
- /; if (seen == false || first`._type == TTYPE_KEYTP || first`.eq("~\0") == true || first`.eq("{\0") == true)
+ /; if (seen == false || first`._type ==TTYPE_KEYTP || first`.eq("~\0") == true || first`.eq("{\0") == true)
_ast_type(fin, ~list, first)
/; if (first`._type !== TTYPE_USRWD)
@@ -1069,8 +1078,102 @@ struct Node {
blf.end()
;/
+# The following code (mhf_post_list and mhf_post)
+# are a convoluted hellscape of terrible
+# design choices. Both functions may recursively call
+# the other, and both produce a multitude of different
+# output values based on what they think the AST should
+# be (sometimes based on the next token, sometimes not).
+
+# For instance, mhf_post tries to return NULL when it
+# thinks the next token means we are in a state where we
+# can parse a declaration (we have seen a valid type and
+# the next token is an identifier) but if we are calling
+# this function from mhf_post_list, we are inside a list,
+# making this an invalid next token in either the case
+# where we are parsing a declaraion OR the case where we
+# are parsing a value. It should blow up, so mhf_post_list
+# checks to see if it's null and returns whatever output
+# list it has created instead so something higher up the
+# chain (mhf_finish_value) can catch it and explode instead.
+# It's almost like an exception. Fuck me.
+
+# Otherwise, mhf_post will return whatever node the mhf_finish_value
+# should use as the node to start stitching together with binary operators
+# and values. Even worse, the node it should return may be a sub-node
+# FROM MHF_POST_LIST!!!
+
+# _mhf_post_list, however, should NEVER return NULL. In the case
+# where it suceeded in parsing a list of types, it will return
+# the passed in ~Node (which SHOULD be a type node), but otherwise,
+# it will return the node it was trying to generate when the failure
+# occurred. That's at least slightly fewer possible options,
+# but it is very important to remember that the token it ends on is
+# the end of the list, which allows checking whether the line
+# number of the next token is the same. If it's not, this WHOLE THING
+# was a FUNCTION CALL, and the ENTIRE FUCKING TREE NEEDS TO BE
+# RE-WORKED THROUGH THE TRANSFORM FUNCTION TO REPRESENT THAT.
+
+# Both functions try to compensate by checking the output
+# state of the other, and it creates a clusterfuck of strange
+# if statements and pointer logic that I wrote in an exremely
+# sleep deprived state. I'm just really hoping this
+# works. It will all probably get refactored later when
+# the standard library is fully flushed out and the compiler
+# is more mature. For now, may this warning be a sufficient
+# deterrant for the common programmar looking to refactor.
+
+# The only saving grace of this is I guess that it's
+
+# If you want to skip the most hellish spaghetti code I have ever
+# had the displesure of writing, just search for "maybe_helper_fun"
+# and have your editor of choice skip to that point.
+
/; _mhf_post_list (~utils.File fin, ~Node mod, ~Token first) [~Node]
- return NULL
+ Node out
+ out.init(NTYPE_TLIST, first`.data)
+ first` = produce_next_token(fin, first`)
+
+ /; loop (bool run = true; run == true && first`._type !== TTYPE_ERR)
+ Node _typ
+ _typ.init(NTYPE_TYPE, utils.strcpy("\0"))
+ _type_helper_pre(fin, ~_typ, first)
+
+ /; if (first`.eq("(\0"))
+ # Oh no! We hit a value!
+ out.add_child(~_typ)
+ return mod`.add_child(~out)
+
+ ;; else if (first`._type == TTYPE_USRWD || first`._type == TTYPE_KEYWD)
+ # Try to keep parsing as a type, if we error out we
+ # will return whatever post returns
+ ~Node cur = _mhf_post(fin, ~_typ, first)
+ /; if (cur == ~_typ && first`.eq(",\0"))
+ # Keep going
+ Token tmp = produce_next_token(fin, first`)
+ first`.end()
+ first` = tmp
+ ;; else if (cur == ~_typ && first`.eq(")\0"))
+ # Stop here, reached end of list
+ run = false
+ ;; else
+ # Some error occurred, mhf_transform will
+ # preserve lists so we return the list we were
+ # creating.
+ # mhf_finish_value can find the last valid parse
+ # before the error occured by getting the last
+ # subnode from the list
+ out.add_child(~_typ)
+ return mod`.add_child(~out)
+ ;/
+ ;/
+
+ out.add_child(_typ)
+ ;/
+
+ mod`.add_child(~out)
+ # Properly parsed type list
+ return mod
;/
/; _mhf_post (~utils.File fin, ~Node mod, ~Token first) [~Node]
@@ -1081,6 +1184,10 @@ struct Node {
id.init(NTYPE_ID, first`.data)
mod`.add_child(~id)
first` = produce_next_token(fin, first`)
+ ;; else if (first`._type == TTYPE_KEYWD)
+ Node id
+ id.init(NTYPE_ID, first`.data)
+ mod`.add_child(~id)
;; else
return mod
;/
@@ -1091,13 +1198,27 @@ struct Node {
first` = tmp
;; else if (first`.eq("(\0") == true || first`._type == TTYPE_USRWD)
run = false
+ ;; else if (first`._type == TTYPE_KEYWD)
+ first` = produce_next_token(fin, first`)
+ run = false
;; else
return mod
;/
;/
/; if (first`.eq("(\0") == true)
- return _mhf_post_list(fin, mod, first)
+ ~Node weird = _mhf_post_list(fin, mod, first)
+ # We get back right at the end paren of the list.
+ # If the next token is a USRWD on the same line,
+ # we want to return NULL, otherwise return the weird
+ # pointer
+ int ln = first`.line
+ /; if (weird == mod && _advance_check(fin, first, ")") == true)
+ /; if (first`._type == TTYPE_USRWD && first`.line == ln)
+ return NULL
+ ;/
+ ;/
+ return weird
;/
/; if (first`._type == TTYPE_USRWD)