Added support for progress bars

[mehtamohit013]

Describe your changes

Added support for flushing out stdout and stderr

Issue ticket number and link

Closes #61 #1080

Checklist before requesting a review

• I have performed a self-review of my code
• I have added thorough tests (when necessary).
• I have added the right documentation in the docstring and changelog (when needed)

---

📚 Documentation preview 📚: https://ploomber-engine--66.org.readthedocs.build/en/66/

[mehtamohit013]

Code used to run:

print("Hello World")

for i in range(5):
    print(i)

from tqdm import tqdm
import time

my_list = list(range(100))

with tqdm(total=len(my_list)) as pbar:
    for x in my_list:
       time.sleep(0.1)
       pbar.update(1)
       if x%20==0:
           print(x)
print("Hello World")
print('Hello World 2')

Output

Issues:

• \r character when tqdm end
• Whenever a something is written to stdout when tqdm is running, there is no \n as sys.stderr.write() doesn't include \n and if I include one, progress bar will not be updated in place

Any suggestions on how to resolve this issue?
Maybe we can add some type of cell tag when tqdm is running and resolve it separately from stderr one

[idomic]

Any suggestions on how to resolve this issue?

Why don't you just output it as is?
You can always parse and wait for those characters but I assume they've implemented a mechanism for that already in tqdm?
Also let's keep the discussion in one place (probably here is better).

[edublancas]

I think for this PR, let's just make the output of ploomber-engine be as close as possible to papermill.

Based on the discussion, it sounds like we just need to add 1) cell separators 2) Displaying standard error 2) flushing standard output/error

Once that's in, we can tackle updating the progress bar inline.

[mehtamohit013]

Based on the discussion, it sounds like we just need to add 1) cell separators 2) Displaying standard error 2) flushing standard output/error

1. What do you mean by cell seperators?
2. stderr is currently displayed
3. Flushing messes up the output. So, I have not used flushing

[idomic]

@mehtamohit013 check if you can ask for review now? I've updated the permissions

[edublancas]

cell separators: https://user-images.githubusercontent.com/43580047/225945963-93552b85-83ef-48ae-9993-fd9acf5ba1b6.png

Executing cell X
Ending cell X

ok so if stderr is displayed and we cannot flush output at the moment, I think what's missing is to fix the CI, right?

[idomic]

3. Flushing messes up the output. So, I have not used flushing

So how are we going to show the output as it comes? I think that's essentially the main difference we have to solve it.

[mehtamohit013]

Why don't you just output it as is? You can always parse and wait for those characters but I assume they've implemented a mechanism for that already in tqdm?

The problem lies with the fact that I cannot make a parser just only for tqdm as there may be other messages pushed to stderr. Currently, I am just simply redirecting the notebook stderr and stdout to the original stderr and stdout

ok so if stderr is displayed and we cannot flush output at the moment, I think what's missing is to fix the CI, right?

No, tqdm messes up with the output. You can see in the image attached above
Code used: self.default.write(s) where self.default is the original stdout and stderr, before it is switched with IO() class

So how are we going to show the output as it comes? I think that's essentially the main difference we have to solve it.

I think it is because of yield. I will look into flushing more

def patch_sys_std_out_err(display_output):
    """Path sys.{stout, sterr} to capture output"""
    # keep a reference to the system ones
    stdout, stderr = sys.stdout, sys.stderr

    # patch them
    stdout_stream = IO(default = stdout,std_type='out', display = display_output)
    stderr_stream = IO(default = stderr, std_type='err')
    sys.stdout, sys.stderr = stdout_stream, stderr_stream

    try:
        yield stdout_stream, stderr_stream
    finally:
        # revert
        sys.stdout, sys.stderr = stdout, stderr

class IO(StringIO):
    def __init__(self, default,std_type,display = True, newline="\n", initial_value="" ):
        super().__init__(initial_value=initial_value, newline=newline)
        self.default = default
        self.std_type = std_type
        self.display = display
        self._values = []

    def write(self, s):
        self._values.append(s)
        if self.display:
            self.default.write(s)
        super().write(s)

    def get_separated_values(self):
        return self._values[:]

[edublancas]

ok, so then the only "easy" thing we can do at the moment are the cell separators right? I'd suggest opening a PR for that, we merge, make a release. and then we investigate this flushing issue

[mehtamohit013]

ok, so then the only "easy" thing we can do at the moment are the cell separators right? I'd suggest opening a PR for that, we merge, make a release. and then we investigate this flushing issue

Yeah sure

Btw this is the output with flushing (Observe the '0', which messes with ploomber progress bar, because ploomber progress bar updates after cell is executed)

[idomic]

I think it is because of yield. I will look into flushing more

Yeah I think you're right, I also assume it works fine: tqdm + papermill?

[mehtamohit013]

Yeah I think you're right, I also assume it works fine: tqdm + papermill?

No, it doesn't 😅

Ours is messy but tqdm progress bar is updated inplace and we are registering \r also.
The robust solution is to make a custom wrapper with tqdm

Below is the output with papermill-2.4.0:

[idomic]

😂😂 I guess it's a nice to have, we should open an issue just for this since it's probably more than a quick fix!

…

On Sat, Mar 18, 2023, 11:24 Mohit Mehta ***@***.***> wrote: Yeah I think you're right, I also assume it works fine: tqdm + papermill? No, it doesn't 😅 Ours is messy but tqdm progress bar is updated inplace and we are registering \r also. The robust solution is to make a custom wrapper with tqdm Below is the output with papermill: [image: Screenshot_2023-03-18_11-19-45] <https://user-images.githubusercontent.com/43580047/226114862-00b101a2-7108-4a19-a434-868a50add0e0.png> — Reply to this email directly, view it on GitHub <#66 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACYPJONQ76Z2PSL5TPHTR7TW4XHTRANCNFSM6AAAAAAV7C4ZNU> . You are receiving this because you commented.Message ID: ***@***.***>

[mehtamohit013]

@edublancas @idomic
If you want a quick fix, how about this: Just some extra \n but the progress bar is working

[idomic]

I think it's a good temporary solution (might be a bit confusing for the users)

…

On Sat, Mar 18, 2023, 11:36 Mohit Mehta ***@***.***> wrote: @edublancas <https://github.com/edublancas> @idomic <https://github.com/idomic> If you want a quick fix, how about this: Just some extra \n but the progress bar is working [image: Screenshot_2023-03-18_11-34-42] <https://user-images.githubusercontent.com/43580047/226115554-8881b4eb-061d-46ea-be46-bc76ede3861e.png> — Reply to this email directly, view it on GitHub <#66 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACYPJONHOKDS627HF4MYSXDW4XJAVANCNFSM6AAAAAAV7C4ZNU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[mehtamohit013]

@edublancas @idomic If you want a quick fix, how about this: Just some extra \n but the progress bar is working

Implemented this one

[idomic]

looks good besides the leading \n .

[idomic]

It doesn't closes ploomber/ploomber#1080 since we're still using papermill there.