Use a fixed-size reader stack

This improves performance, and makes the reader more suitable for embedded or
network-facing applications, at the cost of requiring the user to specify a
maximum stack size.

1 files changed, 9 insertions, 0 deletions

diff --git a/doc/serdi.1 b/doc/serdi.1
index 8747be92..3ae42ea0 100644
--- a/doc/serdi.1
+++ b/doc/serdi.1
@@ -9,6 +9,7 @@
 .Op Fl abefhlqv
 .Op Fl c Ar prefix
 .Op Fl i Ar syntax
+.Op Fl k Ar bytes
 .Op Fl o Ar syntax
 .Op Fl p Ar prefix
 .Op Fl r Ar root
@@ -68,6 +69,14 @@ Case is ignored, valid values are:
 .Dq TriG ,
 .Dq Turtle .
 .Pp
+.It Fl k Ar bytes
+Parser stack size.
+For performance reasons, parsing is performed with a fixed-size stack.
+By default, the stack is one page (4096 bytes), which should be sufficient for parsing most documents.
+This option can be used to increase or decrease the amount of memory available for parsing.
+It must be large enough to fit one statement, with some extra space for internal records.
+Something around twice the length of the longest statement in text (as if it were written in NTriples or NQuads) is a reasonable value.
+.Pp
 .It Fl l
 Lax (non-strict) parsing.
 If this is enabled, recoverable syntax errors will print a warning, but parsing will proceed starting at the next statement if possible.