It is the 2nd step of previous question (https://mcmap.net/q/6056/godot-method-to-convert-png-incl-alpha-into-2-files-in-etc1-format-via-2-uniforms)
What I know regarding texture() and textureFecth() are detailed below.
texture() uses float within the UV coordinates and needs an info such as nearest or linear in openGL.
textureFecth() uses int within the texture/screen coordinates and needs an extra parameter "LOD"
Previous reply (https://mcmap.net/q/5498/godot-texelfetch-equivalent-in-gles2) is available for a similar help request .
Stickler The second function only applies a filter if you setup the sampled texture to do so. Most if not all Godot textures have a filter property. If you disable that along with mipmaps you are guaranteed to have no filters applied to the texture.
Rosenblast Thank you! It worked but texture still has very low accuracy. texelFetch was able to extract values from sampler2D with 32 bit float accuracy
Success seems partial.
I have converted a lot of Shadertoy codes from tectureFetch to texture together with int/float and UV/screen changes in Shadertoy and tranferring Godot. Almost % 50 of the code gives full success in 2 two mediums. others are not only have low accuracy problem and also full logic changes. And alot of missing.
I believe there is a way to implement the multiple texture() functions with a mipmap/LODs (and maybe offsets for linear/bilinear/trilinear) instead of a single texelFetch() command dor a full equivalency.
Maybe an array can work like in "vec4_type texture (sampler2DArray_type s, vec3 uv [, float bias])" with Mipmapped series.
Can you help me please?