38 lines
3.0 KiB
Plaintext
38 lines
3.0 KiB
Plaintext
[video member=pervognsen stream_platform=twitch project=bitwise title="Domain-Specific Languages In Python" vod_platform=youtube id=4gJOiQbhWWM annotator=Miblo]
|
||
[0:09][Recap and set the stage for the day][:speech]
|
||
[0:34][Review some off-stream :optimisation on the lowered jump table-based switch statement][:asm :language :speech]
|
||
[5:30][Being careful of associativity when flattening expressions and register allocations, with another mention of the Sethi–Ullman algorithm[ref
|
||
site=Wikipedia
|
||
page="Sethi–Ullman algorithm"
|
||
url=https://en.wikipedia.org/wiki/Sethi%E2%80%93Ullman_algorithm]][:asm :language :speech]
|
||
[12:00][Setup to cover domain-specific languages in Python][:language :speech]
|
||
[15:46][Dive into building up an abstract syntax tree for our logic design code, starting by :parsing the & operator][:"data structure" :language]
|
||
[20:42][:Run it to find that it does execute][:language]
|
||
[21:11][Parse the | operator and introduce UnaryNode][:language :parsing]
|
||
[24:02][Note why __not__ doesn't exist in Python][:language :speech]
|
||
[25:11][Guard against erroneous boolean comparison of variables][:parsing]
|
||
[25:54][:Run it to see our intended error]
|
||
[27:50][Fix the \~ Node to use __invert__ (rather than __inv__)][:parsing]
|
||
[28:13][Note that we're building up a graph rather than a tree][:parsing :speech]
|
||
[30:43][Introduce the notion of a ConstantNode, and as_node()][:language :parsing]
|
||
[34:30][A few words on Python's reverse operator overload methods][:language :speech]
|
||
[38:18][Rename InputNode to VariableNode and input() to var()][:language :parsing]
|
||
[39:16][Consider how to handle out-of-range errors for bitwise-not in Python][:language :parsing :speech]
|
||
[44:12][Enable as_node() to handle out-of-range errors on ints by taking their least significant bit][:language]
|
||
[45:05][@elavid][I guess "1 & x" tries to run the int class __and__ function first, and when that throws some exception it tries Per's __rand__ function instead][:language]
|
||
[47:27][Determine to create a simple interpreter][:language :parsing :speech]
|
||
[49:09][Embark on a simple interpreter, introducing evaluate() and related functions][:language :parsing]
|
||
[57:02][Test evaluate() to see that it all works as expected][:language :parsing :run]
|
||
[1:01:28][Construct a nested tree 16 levels deep for evaluate() to evaluate][:language]
|
||
[1:06:33][:Run it to see that the :performance of our evaluate() is terrible, on 16 and 32 levels deep][:language :parsing]
|
||
[1:07:47][Introduce eval() to cache values for evaluate() to call][:language]
|
||
[1:12:15][:Run it to see that the state does get filled in][:language]
|
||
[1:12:41][Increase our nested tree to 32 levels deep][:language]
|
||
[1:13:19][:Run it to see that it is now instantaneous][:performance]
|
||
[1:14:51][Reflect on our Python-based interpreter][:speech]
|
||
[1:15:37][Q&A][:speech]
|
||
[1:15:47][@rickylqhan][What's your endgame?]
|
||
[1:17:11][@garryjohanson][@pervognsen So this is off-topic, but do you think the Intel 8086 chip will be within the direct scope of this project, or is it more complicated than what we are attempting?]
|
||
[1:18:17][That's it for today][:speech]
|
||
[/video]
|