Containers for photos, with scroll, drag&drop between them

cvp

@jmv38 During my tests and my multiple crashes, I'm pretty sure I have seen a red message during one microsecond, containing the words Decompression Bombs

This is a [PIL message](Decompression Bombs) .
I didn't use PIL. Perhaps the ui.Image.save uses it...

Something like, it not sure, message disappeared immediately
/usr/lib64/python2.6/site-packages/PIL/Image.py:2261: DecompressionBombWarning: Image size (166109750 pixels) exceeds limit of 89478485 pixels, could be decompression bomb DOS attack. DecompressionBombWarning)

cvp

@jmv38 said:

However it is strange that 15000 is ok bu not 9000...?

15000 is in pixel, already multiplied by 2. The width was 7776, just < 9000

cvp

Same crash with

Image.warnings.simplefilter('error', Image.DecompressionBombWarning)
Image.MAX_IMAGE_PIXELS = 100000000000

jmv38

@cvp 5000 doesnt work for me and 5000<7000

cvp

@jmv38 and Sure that you don't have two pixels py point?
What is your device?

jmv38

@cvp i do have 2 pixels per point (ipad air)
i dont crash, it is just that the image saved is black
i checked that 4x1024 is ok and 4x1025 fails
i remember this ios limit 4096 from somwhere.

cvp

@jmv38 could you post your code, only this part, and the image you use?

jmv38

@cvp just to let you know here is my code.
It wont run because the rest of the code is missing, but it gives you the idea

def save(self):
    # make a hi resolution copy of back & images, then save it in camera roll
    xo, yo, w, h = self.page.back.frame
    c = self.page.back.background_color
    targetWidth = 4*1024
    s = targetWidth / w
    w, h = w*s, h*s
    page = ui.View( frame=(0,0,w,h), background_color=c)
    views = []
    for thumb in self.thumbs:
      x,y,w,h = thumb.frame
      x,y,w,h = (x-xo)*s, (y-yo)*s, w*s, h*s
      v = ui.View( frame=(x,y,w,h) )
      x,y,w,h = thumb.iv.frame
      x,y,w,h = x*s, y*s, w*s, h*s
      img = thumb.getImage(thumb.asset)
      iv = ui.ImageView(frame=(x,y,w,h), image=img)
      v.add_subview(iv)
      page.add_subview(v)
      views.append(v)
    # save page image in pythonista
    getTopView().add_subview(page)
    #page.bring_to_front()
    #if True: return
    path = 'temp.jpg'
    with ui.ImageContext(page.width, page.height) as ctx:
      page.draw_snapshot()                            
      ui_image = ctx.get_image()
    pil = Image.open(io.BytesIO(ui_image.to_png()))
    pil.save(path , quality=99)
    # save page image in albums
    asset = photos.create_image_asset(path)
    os.remove(path)
    getTopView().remove_subview(page)
    views = False
    console.hud_alert('saved')

looks like i must add the view to the screen to get the draw snapshot to work.

cvp

@jmv38 said:

looks like i must add the view to the screen to get the draw snapshot to work.

Not at all, I think. You've seen my little code, nothing goes to the screen

jmv38

@cvp you are correct. My pb was pbly sthg else.

cvp

@jmv38 The only 4096 limit I find for iPad Air is Max OpenGL Texture Sizes

cvp

From here

The UIImage documentation (as of iOS 10) no longer seems to mention size limitations, although if you use UIImageView with images whose dimensions are larger than the maximum texture size* supported by the device you happen to be using you do get very large memory consumption at render time.

(The memory consumption I see in Instruments seems to indicate that the entire image is put into a 32 bits per pixel buffer when the CA::Layer is rendered.)

If your app doesn't get kill by the OS due to memory usage, the UIImageView does still end up displaying the image though.

Given this, you'll still need strategies to deal with very large images.

You can check the maximum texture size using something like glGetIntegerv(GL_MAX_TEXTURE_SIZE, &maxTextureSize);. Just make sure you've set the EAGLContext current context to be something non-nil before querying OpenGL, otherwise you'll get zero.

cvp

I've tried in ObjectiveC, following this and this

and I have same crash with very big files of more than 25MB

# https://gist.github.com/jsbain/389a67c5aacb097b87fd
# https://github.com/tdamdouni/Pythonista/blob/master/_2017/core_image.py
import ui
from objc_util import *
import ctypes
import os

iv = ui.ImageView()
iv.image = ui.Image.named('P1020096.JPG') # 3888 x 2592 pixels
wi,hi = iv.image.size
iv.frame = (0,0,wi,hi)
print(wi,hi)															# 3888 2592
w = 7000
h = w*hi/wi
print(w,h)																# 7776 5184
with ui.ImageContext(w,h) as ctx:
	iv.draw_snapshot()
	ui_image = ctx.get_image()
uio = ObjCInstance(ui_image)
c.UIImageJPEGRepresentation.argtypes = [c_void_p, CGFloat]
c.UIImageJPEGRepresentation.restype = c_void_p
quality = 1.0
data = ObjCInstance(c.UIImageJPEGRepresentation(uio.ptr, quality))
filename = os.path.abspath('t1.jpg')
data.writeToFile_atomically_(filename, True)