From 8b9754ff50fc905b39bcb777bf3f8b08b79c1cbb Mon Sep 17 00:00:00 2001
From: matt rice <ratmice@gmail.com>
Date: Sat, 24 Aug 2024 14:23:16 -0700
Subject: [PATCH] docs: Add a chapter on start states to the book

---
 doc/src/SUMMARY.md          |   1 +
 doc/src/lexcompatibility.md |   4 +-
 doc/src/start_states.md     | 125 ++++++++++++++++++++++++++++++++++++
 3 files changed, 128 insertions(+), 2 deletions(-)
 create mode 100644 doc/src/start_states.md

diff --git a/doc/src/SUMMARY.md b/doc/src/SUMMARY.md
index a1e8d44ef..e142efb09 100644
--- a/doc/src/SUMMARY.md
+++ b/doc/src/SUMMARY.md
@@ -5,6 +5,7 @@
 - [Lexing](lexing.md)
   - [Lex compatibility](lexcompatibility.md)
   - [Hand-written lexers](manuallexer.md)
+  - [Start States](start_states.md)
 - [Parsing](parsing.md)
   - [Yacc compatibility](yacccompatibility.md)
   - [Return types and action code](actioncode.md)
diff --git a/doc/src/lexcompatibility.md b/doc/src/lexcompatibility.md
index 6567a5946..a1d47b02b 100644
--- a/doc/src/lexcompatibility.md
+++ b/doc/src/lexcompatibility.md
@@ -26,8 +26,8 @@ There are several major differences between Lex and grmtools:
  * Both Lex and grmtools lex files support start conditions as an optional prefix
    to regular expressions, listing necessary states for the input expression to 
    be considered for matching against the input. Lex uses a special action
-   expression `BEGIN(state)` to switch to the named `state`. grmtools lex files
-   use a token name prefix.
+   expression `BEGIN(state)` to switch to the named `state`. Start states in grmtools
+   are described in [start_states](start_states.md).
 
  * Character sets, and changes to internal array sizes are not supported by grmtools.
 
diff --git a/doc/src/start_states.md b/doc/src/start_states.md
new file mode 100644
index 000000000..b60dca02e
--- /dev/null
+++ b/doc/src/start_states.md
@@ -0,0 +1,125 @@
+# Start States
+
+The following explains the syntax and semantics of Start States in lrlex.<br>
+A working example can be found in the repository at [lrpar/examples/start_states][1]
+
+[1]: https://github.com/softdevteam/grmtools/tree/master/lrpar/examples/start_states
+## Motivation
+
+Start states are a feature from lex which can be used for context sensitive lexing.
+For instance, they can be used to implement nested comments (see the example in the repository).
+Such that the tokens start/end markers of tokens maintain balance.
+
+This is achieved by making rules which are qualified to match only when the lexer is in a
+particular state. Additionally the lexer has a stack of states, and matching rules perform actions
+which modify the stack.
+
+## The INITIAL start state
+Unless specified otherwise all lex rules are members of the *INITIAL* start state.
+
+```
+%%
+<INITIAL>a "A"
+<INITIAL>[\t \n]+ ;
+```
+
+This is the lex file below with no start states specified.
+
+```
+%%
+a "A"
+[\t \n]+ ;
+```
+
+## Rules matching multiple states
+
+Rules can be matched in multiple states, just separate the states a rule should match in commas.
+The following matches the `a` character when in either of the states `FirstState` or `SecondState`.
+
+```
+<FirstState,SecondState>a "A"
+```
+
+## Differences from POSIX lex
+
+In posix lex start states are entered via code in the action, through either `BEGIN(STATE)` and
+calling combinations of `yy_push_state`, and `yy_pop_state`.
+
+Because lrlex is actionless, and does not support code actions, instead we have operators to
+perform the common modifications to the stack of start states.
+
+### Push
+The push operator is given by the adding '+' to the target state on the right hand side within
+angle brackets. The following when regex matches in the *CURRENT_STATE* pushes *TARGET_STATE* to
+the top of a stack of states.
+
+```
+<CURRENT_STATE>Regex <+TARGET_STATE>;
+```
+
+### Pop
+The pop operator is given by the adding '-' to the target state on the right hand side within angle
+brackets. Similarly when in the current state, the following pops the current state off of the
+stack of states. Similarly to calling `yy_pop_state` from action code.
+```
+<CURRENT_STATE>Regex <-CURRENT_STATE>;
+```
+
+### ReplaceStack
+The ReplaceStack operator is given by naming the target state within angle brackets.
+The ReplaceStack op clears the entire stack of states, then pushing the target state.
+
+```
+<CURRENT_STATE>Regex <TARGET_STATE>;
+```
+
+### Returning a token while performing an operator.
+Start state operators can be combined with returning a token for example:
+
+```
+<CURRENT_STATE>Regex <+TARGET_STATE>"TOKEN"
+```
+
+## Adding a start state
+Start stats come in two forms, *exclusive* and *inclusive*. These are given by `%x` and `%s`
+respectively.
+
+### Exclusive states
+In an exclusive state, the rule can be matched *only* if the rule begins with the state specified.
+In the following because `ExclState` is *exclusive*, the `#=` rule is only matched during the
+`INITIAL` state, while the `a` and `=#` characters are only matched while in the `ExclState`.
+
+```
+%x ExclState
+%%
+
+#= <+ExclState>;
+<ExclState>a "A"
+<ExclState>=# <-ExclState>;
+```
+
+### Inclusive states
+
+Inclusive states are added to the set of rules to be matched when the start state is unspecified.
+
+```
+%s InclusiveState
+%%
+
+a "A"
+<InclusiveState>b "B"
+<INITIAL>#= <+InclusiveState>;
+<InclusiveState>=# <-InclusiveState>;
+```
+
+Is equivalent to the following using exclusive states.
+
+```
+%x Excl
+%%
+
+<INITIAL, Excl>a "A"
+<Excl>b "B"
+<INITIAL>#= <+Excl>;
+<Excl>=# <-Excl>;
+```