todos.md

Jesse

untagged

- #272 Fold `-` sign into this value at tokenization time?
- #154 This is pretty shitty because whenever we copy one of these structs this field has to be manually zeroed out ..
- #262 These should probably be in their own type_ast_node_cast .. ?
- #366 Can this go in the posix platform?  How do we open dlls?
- #367 How do we turn this on?  Currently on my machine it causes
- #343 I think this is busted on invalid program input.
- #278 Disallow invalid suffixes lul/LUL .. LUU .. ULLLL etc..
- #363 A bunch of this crap can get taken out of here..  I added it
- #304 only do this work once
- #282 Should we parse out the bonsai_function def explicitly here?
- #345 How do we differentiate between arithmetic/logical shifts here?
- #299 This could be improved by not taking the StructName, and
- #260 Currently ignoring the second half of a ternary .. we should probably not do this
- #264 Once we have proper macro expansion, this can be expanded and concatenated to the string as well.
- #316 These paths are exactly the same .. they should probably use the same API ..?
- #340 We should only traverse files that were passed to us on the CLI
- #265 Unnecessary .. I just added these as a hack get parsing to work
- #266 Unnecessary .. I just added it as a hack get parsing to work
- #269 Unnecessary .. I just added these as a hack get parsing to work
- #270 Unnecessary .. I just added these as a hack get parsing to work
- #271 Unnecessary .. I just added these as a hack get parsing to work
- #267 Unnecessary .. I just added these as a hack get parsing to work
- #268 Unnecessary .. I just added these as a hack get parsing to work
- #368 How do we get vsync here?
- #275 Rewrite this in terms of `ToU32(counted_string S)`

immediate

- #238 Change name to BitwiseAnd
- #239 Change name to BitwiseOr
- #347 Nuke this
- #83 d_union-ify this
- #348 This should go into an AreEqual bonsai_function I think..
- #296 When we properly traverse include graphs, this assert should not fail.
- #222 Re-add [[nodiscard]] here

high_priority

- #101 Profile and check collision rate of this!
- #185 these should be printable!
- #182 This should respect Args.Outpath passed in!
- #183 This should respect Args.OutPath passed in!
- #230 ParseVariable should work with bonsai_function pointer types

format_counted_string

- #91 Jettison this .. Can it be metaprogrammed?

compiler_feature

- #99 Generate this?  Need a compiler feature to generate stuff from primitive types.

memory_consumption

- #85 Compress to 16 bit float?
- #86 Unneeded for projectiles. factor out of here?
- #138 32F is only necessary for reprojection of Position for calculating AO.  Consider passing this in when creating a Texture?

speed

- #87 Re-enable this
- #92 This could be optimized significantly
- #96 Vectorize for speed boost!
- #97 Vectorize for speed boost!
- #101 Profile and check collision rate of this!
- #106 Pass this in!
- #107 Can we compute this with MouseP to avoid a frame of input delay?
- #295 This could be optimized significantly by shuffling the logic around, not to mention using hashtables.
- #120 Allocate lights such that this swizzle is unneeded
- #127 Pretty sure we don't have to set the faces anymore??
- #128 Do culling on these as well?
- #129 Can we somehow atomically pull this one off the queue and initialize it on demand?
- #132 Pretty sure we can do some XOR trickery or something here to avoid this branch, which could be a large perf win
- #136 Why are these allocated on the heap?  Seems unnecessary..
- #139 Is this necessary to avoid some pointer aliasing bug?
- #364 This should probably go away and make sure we always just use counted strings
- #365 This should probably go away and make sure we always just use counted strings

cleanup

- #94 Make this non-const?
- #77 Axe this!
- #238 Change name to BitwiseAnd
- #239 Change name to BitwiseOr
- #188 This should have the name property, instead of having the struct and enum defs have seperate names. * Actually .. it's unclear if this is true or not anymore since we've added a bunch more stuff to the union.
- #88 Perlin.h depends on this .. rewrite it.
- #89 Perlin.h depends on this .. rewrite it.
- #90 Redistribute this
- #91 Jettison this .. Can it be metaprogrammed?
- #103 Replace this expression everywhere!!!  This might be done.
- #104 Test this actually gets respected!!
- #105 Test this actually gets respected!!
- #106 Pass this in!
- #107 Can we compute this with MouseP to avoid a frame of input delay?
- #108 Do we have to pass the style to both of these functions, and is that a good idea?
- #109 Reset this in here?
- #113 Remove these?
- #144 This actually has nothing to do with the platform

rng

- #82 257 seemed to produce slightly worse results, but the source literature seemed to indicate using a constant close to the total number of discrete characters you'll encounter is advisable.  I believe this to be somewhat higher than 53, but it would be worth testing on real-world data (source files).
- #95 This is LCG RNG - do we want a better one?

metaprogramming

- #191 This can be generated, but it requires lazily evaluating functions once their dependant types have been generated.  Not sure if this is a good idea or not, but we could do it. * meta( generate_stream_struct(struct_member) ) * Currently the invariant of being able to remove and re-generate all meta output in a single run of the metaprogramming system is broken by this.
- #290 generating this: meta( d_union declaration { function_decl variable_decl }) results in a name collision with the struct_member union tag. * Should we have some way of overriding the tag name it generates?  Or potentially modify the way we generate type tags such that name collisions won't happen.  I'd prefer an override to keep the tag names as concise as possible, but maybe once the preprocessor generates the switch statements for us it won't matter if they're overly verbose.
- #193 Metaprogram this.  I've had bugs multiple times because of it.
- #83 d_union-ify this
- #91 Jettison this .. Can it be metaprogrammed?
- #99 Generate this?  Need a compiler feature to generate stuff from primitive types.
- #100 Do we actually want this in here?
- #301 These functions are super repetitive, generate them!
- #226 Should we handle this differently?
- #222 Re-add [[nodiscard]] here
- #230 ParseVariable should work with bonsai_function pointer types
- #121 Need a way of dynamically allocating more of these on demand if this unlink fails!
- #236 Metaprogram a free_mesh_stream!
- #237 Metaprogram this!!
- #160 Generate this!

robustness

- #78 How should we actually set this?
- #82 257 seemed to produce slightly worse results, but the source literature seemed to indicate using a constant close to the total number of discrete characters you'll encounter is advisable.  I believe this to be somewhat higher than 53, but it would be worth testing on real-world data (source files).
- #85 Compress to 16 bit float?
- #98 Make allocating these on the stack work!
- #104 Test this actually gets respected!!
- #105 Test this actually gets respected!!
- #109 Reset this in here?
- #121 Need a way of dynamically allocating more of these on demand if this unlink fails!
- #122 Actually read all the data!
- #124 Load models in multiple chunks instead of one monolithic one. The storage for chunks must be as large as the largest chunk we will EVER load, which should definately not be decided at compile time.
- #125 Are we really done?
- #128 Do culling on these as well?
- #135 Query max gpu textures?
- #137 This _should_ be able to be glTexImage3D, but the driver is throwing an error .. why?!
- #139 Is this necessary to avoid some pointer aliasing bug?
- #143 This could probably be made better by writing to a statically allocated buffer ..?

font

- #77 Axe this!
- #140 Can we support these?

over_allocation

- #78 How should we actually set this?
- #79 Can this just be a v3?
- #93 This seems like _way_ overkill..
- #148 Better or more accurate way of allocating this size?
- #149 Better way of allocating this?

ummmm

- #79 Can this just be a v3?

xml

- #80 This could be factored out of here if xml_tag::Properties thing was a xml_token_stream* - which would require a two-pass parser most likely.  ie. Tokenizer and AST builder

back_burner

- #161 Reinstate this!
- #80 This could be factored out of here if xml_tag::Properties thing was a xml_token_stream* - which would require a two-pass parser most likely.  ie. Tokenizer and AST builder
- #117 Transient arena for this instead of stack allocation ?

string_hash

- #81 Theres' a null terminated string-hash bonsai_function that should be replaced by this one most likely
- #101 Profile and check collision rate of this!

hashing

- #82 257 seemed to produce slightly worse results, but the source literature seemed to indicate using a constant close to the total number of discrete characters you'll encounter is advisable.  I believe this to be somewhat higher than 53, but it would be worth testing on real-world data (source files).
- #147 Profile this and see if it's reasonable

opengl

- #84 This only gets used when computing the shadow map, so I'm not even sure if it works ATM
- #119 Not present on ES2 .. should we use them?
- #135 Query max gpu textures?
- #138 32F is only necessary for reprojection of Position for calculating AO.  Consider passing this in when creating a Texture?
- #150 Should this be done elsewhere?

completeness

- #84 This only gets used when computing the shadow map, so I'm not even sure if it works ATM
- #140 Can we support these?

engine

- #84 This only gets used when computing the shadow map, so I'm not even sure if it works ATM

entity

- #86 Unneeded for projectiles. factor out of here?

cache_friendly

- #87 Re-enable this

perlin

- #88 Perlin.h depends on this .. rewrite it.
- #89 Perlin.h depends on this .. rewrite it.

aabb

- #92 This could be optimized significantly

open_question

- #94 Make this non-const?
- #100 Do we actually want this in here?
- #125 Are we really done?
- #137 This _should_ be able to be glTexImage3D, but the driver is throwing an error .. why?!
- #142 Is there a better way of making that adjustment?
- #143 This could probably be made better by writing to a statically allocated buffer ..?
- #151 Not getting vsync on my arch laptop.

improvement

- #290 generating this: meta( d_union declaration { function_decl variable_decl }) results in a name collision with the struct_member union tag. * Should we have some way of overriding the tag name it generates?  Or potentially modify the way we generate type tags such that name collisions won't happen.  I'd prefer an override to keep the tag names as concise as possible, but maybe once the preprocessor generates the switch statements for us it won't matter if they're overly verbose.
- #95 This is LCG RNG - do we want a better one?
- #102 Write actual/working fModf!

api_improvement

- #98 Make allocating these on the stack work!

profile

- #101 Profile and check collision rate of this!

hash

- #101 Profile and check collision rate of this!

already_done_elsewhere

- #101 Profile and check collision rate of this!

math

- #102 Write actual/working fModf!

potential_bug

- #108 Do we have to pass the style to both of these functions, and is that a good idea?
- #123 Should these 0s be set to s32_MIN??

bootstrap_debug_system

- #112 This is copy-pasted from the callgraph tests .. should we be able to call this from anywhere?  It's also in the platform layer

copy_paste

- #112 This is copy-pasted from the callgraph tests .. should we be able to call this from anywhere?  It's also in the platform layer

transient_memory

- #117 Transient arena for this instead of stack allocation ?
- #118 Transient storage?
- #134 Transient storage
- #141 Temp-memory?

allocation

- #117 Transient arena for this instead of stack allocation ?
- #120 Allocate lights such that this swizzle is unneeded
- #133 Allocate in a more sensible way?
- #136 Why are these allocated on the heap?  Seems unnecessary..

es2

- #119 Not present on ES2 .. should we use them?

vox_loader

- #122 Actually read all the data!
- #123 Should these 0s be set to s32_MIN??
- #124 Load models in multiple chunks instead of one monolithic one. The storage for chunks must be as large as the largest chunk we will EVER load, which should definately not be decided at compile time.
- #125 Are we really done?

set_to_f32_min

- #123 Should these 0s be set to s32_MIN??

dead_code

- #127 Pretty sure we don't have to set the faces anymore??

correctness

- #128 Do culling on these as well?

theading

- #129 Can we somehow atomically pull this one off the queue and initialize it on demand?

gameplay_improvement

- #129 Can we somehow atomically pull this one off the queue and initialize it on demand?

be_smarter

- #130 This offset is only used to check if entities are grounded.  Can we do that in a more intelligent way?

not_implemented

- #131 This path needs to call CanBuildWorldChunkMesh or something similar
- #132 Pretty sure we can do some XOR trickery or something here to avoid this branch, which could be a large perf win
- #133 Allocate in a more sensible way?

texture

- #135 Query max gpu textures?

ttf_rasterizer

- #140 Can we support these?
- #141 Temp-memory?

format_counted_string_api

- #143 This could probably be made better by writing to a statically allocated buffer ..?

platform

- #144 This actually has nothing to do with the platform
- #150 Should this be done elsewhere?

platform_linux

- #151 Not getting vsync on my arch laptop.

gross

- #52 Gross..

hot_reload

- #153 Doing this properly requires some extra-credit work first.

debug_lib

- #153 Doing this properly requires some extra-credit work first.

hashtable

- #160 Generate this!

debug_recording

- #161 Reinstate this!

ui

- #110 Should we do something special when interacting with this thing instead of Ignored-ing it?

semantic

- #110 Should we do something special when interacting with this thing instead of Ignored-ing it?

parsing

- #213 There is a degenerate case here, what if the file ends without a newline? While we're at it, add tests that make sure these functions return sane stuff when files end with comments!
- #214 There is a degenerate case here, what if the file ends with a malformed comment? While we're at it, add tests that make sure these functions return sane stuff when files end with comments!
- #222 Re-add [[nodiscard]] here
- #230 ParseVariable should work with bonsai_function pointer types

bug

- #185 these should be printable!
- #213 There is a degenerate case here, what if the file ends without a newline? While we're at it, add tests that make sure these functions return sane stuff when files end with comments!
- #214 There is a degenerate case here, what if the file ends with a malformed comment? While we're at it, add tests that make sure these functions return sane stuff when files end with comments!

needs_tests

- #213 There is a degenerate case here, what if the file ends without a newline? While we're at it, add tests that make sure these functions return sane stuff when files end with comments!
- #214 There is a degenerate case here, what if the file ends with a malformed comment? While we're at it, add tests that make sure these functions return sane stuff when files end with comments!

todos

- #225 Rewrite with string_from_parser

output

- #226 Should we handle this differently?

easy

- #225 Rewrite with string_from_parser

id_104

- #105 Test this actually gets respected!!

memory_leak

- #350 
- #351 
- #264 This should use a string builder

big

- #293 Constant expression evaluation for #if statements
- #291 Macro functions
- #292 Include graph traversal

id_301

- #302 

id_320

- #321 Once this path goes away, the assertion with the label associated with this todo id should be put back in.
- #319 Type-checking failed.

id_321

- #323 

id_345

- #346 

id_347

- #348 This should go into an AreEqual bonsai_function I think..
- #349 Change to a bonsai_function call