On Mon, Feb 9, 2015 at 6:53 AM, Edward K. Ream <[email protected]> wrote:
> There is no guarantee that this scheme will work, but I have high
hopes. During testing, the new code will be enabled by a new_shadow switch
at the start of leoShadow.py.
Everything appears to work as of rev fff0a5e! From the checkin log:
QQQ
All unit tests pass with new_shadow = True.
However, this push sets new_shadow = False so that only the adventurous
need test the code.
QQQ
>
The 'replace' opcode poses a question that has been swept under the rug.
When replacing one set of lines by another, where should sentinels be
written?
The simple code given earlier will work, but it causes a unit test to
fail. Yes, the test itself could be changed, but the following code
intersperses changed lines as before::
tag,ai,aj,bi,bj = opcode
b_lines = list(reversed(x.b[bi:bj]))
for i in range(ai,aj):
x.put_sentinels(i)
if b_lines:
x.put_plain_line(b_lines.pop())
# Put any trailing lines.
while b_lines:
x.put_plain_line(b_lines.pop())
Imo, the present code is the way the update algorithm is written in The
Book.
Edward
P.S. There was one complication. The 'insert' code now starts with::
x.put_sentinels(ai)
So now all opcode handlers start with a call to put_sentinels(ai), ensuring
that all sentinels get called at *least* once.
A kludge: x.put_sentinels clears x.sentinels[i] after writing it, thereby
ensuring that sentinels get written at *most* once.
You might call this Post Script a proof of the correctness of the
algorithm...
EKR
--
You received this message because you are subscribed to the Google Groups
"leo-editor" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/leo-editor.
For more options, visit https://groups.google.com/d/optout.
