[Request] llGetAssetDetails(key UUID, integer asset_type) | Voters

[Request] llGetAssetDetails(key UUID, integer asset_type)

Journey Bunny

This function returns public metadata about an asset, and it is an essential component of scripting in Second Life! How essential?
 This functionionalty is not something we've lived without for 20+ years; instead, it is so critical that we have continuously needed to use store extra data and use larger LSL scripts--and more sim memory--to get around this missing function.
These are each so essential that we all must constantly jump through data storing and loading hoops constantly to accomplish things like:
Sound:
 Trivially, every game might eg, play a game-start sound, then loop the game's tune, then play the game over sound). Game controls etc might be tied to the timing of the music. But you can't do this in Second Life scripting without knowing the length of the sound files. We do silly things like notecards with the length info and we increase the size of a simple music-playing script significantly
 by adding dataserver handling. (And we lose a lot of flexibility to swap sounds, since we have to hard-code file sizes)
Images:
 Re-texturing a surface and getting the aspect ratio and scale correct--again, we need to write down the size (width and height) of the asset. Particularly with PBR's different uv setup, we need to know if we have non-square images!
Animations:
 Just like sounds, completely essential to know their length in order to string together sequences dynamically. Essential but unknowable, so we again hard-code or write data-loader scripts.
Useage:
So, I would like to see llGetAssetDetails return a list of elements related to the asset-type requested. For example, llGetAssetDetails(TEXTURE_PLYWOOD , ASSET_TEXTURE) would return a list of [512,512]
Sound files would have the sound length
Animation files would have play length (and maybe priority?)
Benefit and consideration:
The advantage to this system: if there's ever sufficient justification to add more details for an asset, tack them onto the list, it's pretty backwards-compatible. If there's justification to add another asset that has useful metadata, add an integer type.
This data is needed but even better it's already not
 sensitive. Ironically, anyone and everyone can have it anywhere except Second Life. Eg, it is trivial to use the CDN api to fetch this info. (I just fetched the width and height of the Plywood texture in the example above using curl). The http request functionality in lsl cannot do this: the simulator blocks incoming datatypes that aren't text--eg, all the media files--with a 499 error and empty body. So, the alternative would be to run a proxy service that does this for people but...if the data is publicly available (it is) then why can't these pieces of metadata be used in-world? It would massively simplify dance/music/animesh/gaming etc.
Future:
Other assets might be interesting to fetch basic data about, like Notecard size (useful to know how many bytes before you try loading it). I think by establishing this function and having a signature that accepts UUID and which "type" of asset, we establish a flexible future-minded capability where any new asset metadata that can be proven to be both useful and
 safe to be known by all could be added without breaking past content.
Technical challenges:
Textures pre-2009 have the w and h written a little differently. The function, under the hood, would invoke 2 different methods. Sounds have the sampling rate and number of samples present, so the length is calculated rather than written in the metadata.
I am unable to fit code for sound and textures (which I already have tested in Python) into this request, so...I'll reply to myself so you can see!

11 hours ago

Journey Bunny

Here's a look at the working, publicly available metadata today (Sorry, I'm on a Python kick)

Code for texture

We only need the header:

curl.exe -s --range 0-255 "<CDN ENDPOINT URL>/?texture_id=89556747-24cb-43ed-920b-47caed15465f" -o header.bin

After fetching header:

import struct
data = open("header.bin", "rb").read()
print("First 52 bytes (SOC + SIZ segment):")
for i in range(0, 52, 16):
    chunk = data[i:i+16]
    hex_part = " ".join(f"{b:02x}" for b in chunk)
    asc_part = "".join(chr(b) if 32 <= b < 127 else "." for b in chunk)
    print(f"  {i:04x}  {hex_part:<47}  |{asc_part}|")
print()
print("Annotated:")
print(f"  [0000-0001]  {data[0:2].hex()}          FF 4F = SOC marker")
print(f"  [0002-0003]  {data[2:4].hex()}          FF 51 = SIZ marker")
print(f"  [0004-0005]  {data[4:6].hex()}        Lsiz = length = {struct.unpack_from('>H', data, 4)[0]}")
print(f"  [0006-0007]  {data[6:8].hex()}        Rsiz = capabilities")
print(f"  [0008-000b]  {data[8:12].hex()}    Xsiz = WIDTH  = 0x{struct.unpack_from('>I', data, 8)[0]:08x} = {struct.unpack_from('>I', data, 8)[0]} px")
print(f"  [000c-000f]  {data[12:16].hex()}    Ysiz = HEIGHT = 0x{struct.unpack_from('>I', data, 12)[0]:08x} = {struct.unpack_from('>I', data, 12)[0]} px")

Output:

First 52 bytes (SOC + SIZ segment):
  0000  ff 4f ff 51 00 2f 00 00 00 00 02 00 00 00 02 00  |.O.Q./..........|
  0010  00 00 00 00 00 00 00 00 00 00 02 00 00 00 02 00  |................|
  0020  00 00 00 00 00 00 00 00 00 03 07 01 01 07 01 01  |................|
  0030  07 01 01 ff 52 00 0c 00 00 00 05 01 05 04 04 00  |....R...........|
Annotated:
  [0000-0001]  ff4f          FF 4F = SOC marker
  [0002-0003]  ff51          FF 51 = SIZ marker
  [0004-0005]  002f        Lsiz = length = 47
  [0006-0007]  0000        Rsiz = capabilities
  [0008-000b]  00000200    Xsiz = WIDTH  = 0x00000200 = 512 px
  [000c-000f]  00000200    Ysiz = HEIGHT = 0x00000200 = 512 px

Journey Bunny

Code for Sound

Little trickier, we can't just get the head. So:

import struct, urllib.request
def get_sound_duration(uuid):
    url = f"<CDN ENDPOINT URL>/?sound_id={uuid}"
    # Request 1 — head bytes, also tells us file size
    req = urllib.request.Request(url, headers={"Range": "bytes=0-63"})
    with urllib.request.urlopen(req) as r:
        head = r.read(64)
        content_range = r.headers.get("Content-Range", "")
        # "bytes 0-63/198176" → split on "/" → 198176
        F = int(content_range.split("/")[1])
    num_segs    = head[26]
    v           = 27 + num_segs
    sample_rate = struct.unpack_from("<I", head, v + 12)[0]
    # Request 2 — tail window, now that we know F
    tail_start  = F - 4096
    tail_end    = tail_start + 2047
    req = urllib.request.Request(url, headers={"Range": f"bytes={tail_start}-{tail_end}"})
    with urllib.request.urlopen(req) as r:
        tail = r.read(2048)
    pos = tail.rfind(b'OggS')
    granule = struct.unpack_from("<q", tail, pos + 6)[0]
    duration = granule / sample_rate
    return duration, sample_rate, F
print(get_sound_duration("df634a11-f28e-371c-404b-1635aee9aadc"))

Output:

29.0, 44100, 198176