Difference between metadata sizes and toBuffer sizes of image ?

Question about an existing feature

What are you trying to achieve?

I am confused in the sizes of image generated after metadata() and toBuffer(). what is the differnce between the sizes of images. I have provided the sample code with outputs, you can see the difference in sizes of image

Please provide a minimal, standalone code sample, without other dependencies, that demonstrates this question

const file = fs.readFileSync("sample.png")
output: - 
<Buffer 89 50 4e 47 0d 0a 1a 0a 00 00 00 0d 49 48 44 52 00 00 13 88 00 00 13 88 08 03 00 00 00 6a 46 77 f9 00 00 03 00 50 4c 54 45 2f 1f 1d 2a 1a 19 29 19 18 ... 1676953 more bytes>

const sharpMetadata = await sharp(file).metadata()
output :-
{
  format: 'png',
  size: 1677003,
  width: 5000,
  height: 5000,
  space: 'srgb',
  channels: 4,
  depth: 'uchar',
  density: 72,
  isProgressive: false,
  paletteBitDepth: 8,
  hasProfile: false,
  hasAlpha: true
}


const {data, info} = await sharp(file).toBuffer({ resolveWithObject: true });
output : - 
<Buffer 89 50 4e 47 0d 0a 1a 0a 00 00 00 0d 49 48 44 52 00 00 13 88 00 00 13 88 08 06 00 00 00 5d 98 87 cb 00 00 00 09 70 48 59 73 00 00 0b 13 00 00 0b 13 01 ... 3270296 more bytes> {
  format: 'png',
  width: 5000,
  height: 5000,
  channels: 4,
  premultiplied: false,
  size: 3270346
}

[g-abhishek]

As you said ‘A call to toFile or toBuffer will almost always generate different output compared with the input, so should be treated as a “lossy” operation. Whether the output is larger or smaller depends entirely upon the processing operations/options you’re using.’

But what if I don’t perform any operation and simply create the sharp image and then create file of that image, then how to achieve the same input size of the image in that file?

[lovell]

https://sharp.pixelplumbing.com/api-output#png

Set palette to true for slower, indexed PNG output.

Indexed PNG images that use a quantised palette are lossy.

https://en.wikipedia.org/wiki/Portable_Network_Graphics#Lossy_PNG_compression