PDF data stream with python

Question

Context: My code fetches a set of coordinates from some png documents and later on performs some redaction in certain fields (it uses these coordinates for drawing rectangles in certain areas).

I want my final output to be a pdf with each redacted image as page. I can achieve this with fpdf package with no problem.

However, I intend to send this pdf file as email (base64 encoded) attachment. Is there any way to get the base64 string from fpdf output?

On top of that, can I use image binary string in fpdf image method?

See the redact_pdf method below (I placed some comments there to be more clear)

Code:

class Redaction:
def __init__(self,png_image_list,df_coordinates):
    self.png_image_list = png_image_list
    self.df_coordinates = df_coordinates
    
def _redact_images(self):
    redacted_images_bin = []
    for page_num,page_data in enumerate(self.png_image_list):
        im_page = Image.open(io.BytesIO(page_data))
        draw = ImageDraw.Draw(im_page)
        df_filtered = self.df_coordinates[self.df_coordinates['page_number'] == page_num+1]
        for index, row in df_filtered.iterrows():
            x0 = row['x0'] * im_page.size[0]
            y0 = row['y0'] * im_page.size[1]
            x1 = row['x1'] * im_page.size[0]
            y1 = row['y1'] * im_page.size[1]
            x2 = row['x2'] * im_page.size[0]
            y2 = row['y2'] * im_page.size[1]
            x3 = row['x3'] * im_page.size[0]
            y3 = row['y3'] * im_page.size[1]
            coords = [x0,y0,x1,y1,x2,y2,x3,y3]
            draw.polygon(coords,outline='blue',fill='yellow')
        redacted_images_bin.append(im_page)
        
    return redacted_images_bin

def redacted_pdf(self):
    redacted_images = self._redact_images()
    pdf = FPDF()
    pdf.set_auto_page_break(0)
    for index,img_redacted in enumerate(redacted_images):
        img_redacted.save(f"image_{index}.png")
        pdf.add_page()
        pdf.image(f"image_{index}.png",w=210,h=297)
        os.remove(f"image_{index}.png") # I would like to avoid file handling!
    pdf.output("doc.pdf","F") # I would like to avoid file handling!
    #return pdf #this is what I want, to return the pdf as base64 or binary

Answer 1

In documentation I found that you can get PDF as string using

pdf_string = pdf.output(dest='S')

so you can use standard module base64

import fpdf
import base64

pdf = fpdf.FPDF()

# ... add some elements ...

pdf_string = pdf.output(dest='S')
pdf_bytes  = pdf_string.encode('utf-8')

base64_bytes  = base64.b64encode(pdf_bytes)
base64_string = base64_bytes.decode('utf-8')

print(base64_string)

Result:

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

---

As for image(): it needs filename (or url) and it can't work with string or io.BytesIO().

Eventually you may get source code and you can try to change it.

There is even request on GitHub: Support for StringIO objects as images

---

EDIT:

I found that there is fork fpdf2 which can use pillow.Image in image() - see fpdf2 Image

And in source code I found image() can also work with io.BytesIO()

---

Example code for fpdf2 (output() gives bytes instead of string)

import fpdf
import base64
from PIL import Image
import io

#print(fpdf.__version__)

pdf = fpdf.FPDF()

pdf.add_page()

pdf.image('lenna.png')

pdf.image('https://upload.wikimedia.org/wikipedia/en/7/7d/Lenna_%28test_image%29.png')

f = open('lenna.png', 'rb')
pdf.image(f)

f = Image.open('lenna.png')
pdf.image(f)

f = open('lenna.png', 'rb')
b = io.BytesIO()
b.write(f.read())
pdf.image(b)

# save in file
pdf.output('output.pdf')

# get as bytes
pdf_bytes = pdf.output()

#print(pdf_bytes)

base64_bytes  = base64.b64encode(pdf_bytes)
base64_string = base64_bytes.decode('utf-8')

print(base64_string)

Wikipedia: Lenna [image]

---

Test for writing in fpdf2

import fpdf   

pdf = fpdf.FPDF()

pdf.add_page()
pdf.image('https://upload.wikimedia.org/wikipedia/en/7/7d/Lenna_%28test_image%29.png')

# --- test 1 ---

pdf.output('output-test-1.pdf')

# --- test 2 ---

pdf_bytes = pdf.output()

with open('output-test-2.pdf', 'wb') as f:  # it will close automatically
    f.write(pdf_bytes)

# --- test 2 ---

pdf_bytes = pdf.output()

f = open('output-test-3.pdf', 'wb')
f.write(pdf_bytes)
f.close()  # don't forget to close when you write