Partial rendering by zanjonke · Pull Request #167 · Codeplain-ai/codeplain

zanjonke · 2026-04-21T09:17:50Z

Summary

Adds partial rendering — when re-running codeplain on a module that has been previously rendered, detect what has changed and offer to resume from the right checkpoint instead of re-rendering everything from scratch.

Given a module tree (e.g. root → middle → leaf), on startup the tool now:

Walks the tree looking for: (a) the last FRID successfully rendered, (b) any module whose plain spec has changed since its last render, (c) any module whose required modules' code has changed since.
Picks the earliest-affected module and presents a TUI with the available continuation choices (continue from next FRID, re-render from start of changed module, re-render everything, quit).
Feeds the user's choice into ModuleRenderer, which renders the chosen starting point before the root module.

The PR can be split in 3 parts:

Refactoring of plain_modules.py and module_renderer.py
Implementation of partial rendering
Addiiton of unit tests

~~TODO: this is waiting on the design from @kaja-s~~ DONE

Here are some preliminary screenshots.

kaja-s · 2026-04-22T08:17:11Z

Here are the final designs for all 13 situations with adjusted content: https://www.figma.com/design/7o1Ql87jRiT7k1Ne0Inugl/design?node-id=11155-2090

pedjaradenkovic

@kaja-s @zanjonke

Before going into reviewing the whole code I did testing. I have 2 modules chickens and hello_world_python that was changed to contain 4 functionalities.

I run codeplain chickens. I rendered 3 functionalities of hello world module and stoped and run the renderer again.

It detected that it can continue rendering from functionality 4 and it gave me that option but I think it should say in which module. Should it say: Continue from functionality 4 in hello_world_python instead?

Module status is IMO slightly confusing. I would rather state something: last rendered functionality or something similar.
Current shortened version of functionality is too short. I suggest we print a couple of lines before concatenating it.
When the whole module is rendered and I start the render of the module again, because it's a very destructive action, I suggest we give a choice to users - start from scratch or quit.
When both modules are fully rendered and I changed something in base module (hello_world) it started implementing everything from scratch without asking me. I think we should ask for that.

kaja-s · 2026-05-12T09:09:18Z

Thanks for the feedback, @pedjaradenkovic.

I agree about adding the module name. I created a v4 in Figma where I made the changes: https://www.figma.com/design/7o1Ql87jRiT7k1Ne0Inugl/design?node-id=11460-1275&t=VB2vlpLfpoT36GZl-4
I didn't want to make too many changes to the partial-rendering screen language from the status TUI screen I think we should, for now, keep the language the same on both screens, and on the status TUI, we use "module status".

We can print the first three and then have an option to expand in parentheses.

If I understand correctly, the points 4. and 5. have already been implemented by Zan.

pedjaradenkovic

Here are some more testing notes:

The modules are rendered in this order:
hello_world -> chicken -> eggs

hello_world module was fully rendered by running codeplain hello_world.plain. After that I run codeplain eggs.plain that should be able to continue rendering chicken and eggs modules but I got this:

There was not change in the specs since the module was rendered. I think this is a bug.

Also, module status is not correct. Eggs was not rendered at all and here it's written fully rendered.

When I press quit I get:

✗ rendering failed

  render id:                    f3a14feb-4a24-4aa1-add0-eb1434be7767
  input file:                   hello_world_python.plain
  generated code folder:        -

functionalities  0  used credits  0  render time  0s

We don't have to fix this as a part of this PR - we can sync with @NejcS and do it separately.

Left some comments but will continue tomorrow.

pedjaradenkovic · 2026-05-12T15:58:12Z

        with open(metadata_path, "w", encoding="utf-8") as f:
            json.dump(module_metadata, f, indent=4)
+
+    def _ensure_module_folders_exist(self, first_render_frid: str, render_conformance_tests: bool):


This method and _ensure_frid_commit_exists and _ensure_previous_frid_commits_exist were copied here but not removed from ModuleRenderer (and they are never used there).

pedjaradenkovic · 2026-05-12T16:02:17Z

+        render_range = plain_spec.get_render_range_from(next_frid, next_module.plain_source)
+        msg = "Continue from"
+        if next_frid != plain_spec.get_first_frid(next_module.plain_source):
+            msg += f" functionality [#5593FF]{next_frid}[/] in module [#5593FF]{next_module.module_name}[/]"


I suggest we remove everything that is related to presentation layer from this class. IMO, this should stay business logic only, return apropriate data structure that is further shown on TUI with appropriate language and colors. WDYT?

pedjaradenkovic · 2026-05-12T16:12:08Z

+                ):
+                    sys.exit(0)
+
+    if partial_render_choice is not None and render_range is not None:


I don't think this change is needed as checks render_range is None in previous statement.

This was added defensively since the following code always expects for one or the other to be not None. But their default values are that exactly.

pedjaradenkovic · 2026-05-12T16:14:29Z

+    render_range: list[str] | None = None
+    msg: str | None = None
+    wipe_later_modules: bool = False
+    is_desctructive: bool = False


Typo: is_destructive

pedjaradenkovic · 2026-05-12T16:15:04Z

+                if partial_render_choice is None or (
+                    partial_render_choice.module is None
+                    and partial_render_choice.render_range is None
+                    and partial_render_choice.msg == "Quit"


Extract strings to constants.

pedjaradenkovic · 2026-05-12T16:30:06Z

+            continue
+
+        module_metadata = _module.load_module_metadata()
+        previous_module = _module.required_modules[-1]


Hm, what it there were multiple required modules? I think this will check only the first one and ignore the rest.

The source code for the current module is always taken from the last in the requires list. If any previous module code changes it will be captured before this.

pedjaradenkovic · 2026-05-12T16:47:03Z

+
+
+@dataclass
+class PartialRender:


I think Partial is giving a wrong impression of what this class represents. I would rename it.

zanjonke · 2026-05-13T07:24:57Z

@pedjaradenkovic

Here are some more testing notes:

The modules are rendered in this order:
hello_world -> chicken -> eggs
hello_world module was fully rendered by running codeplain hello_world.plain. After that I run codeplain eggs.plain that should be able to continue rendering chicken and eggs modules but I got this:

I couldnt manage to replicate this behaviour using airplain. I suggest we try to resolve this in person.

pedjaradenkovic

Leaving the comments to follow up on them on Monday.

pedjaradenkovic · 2026-05-13T06:45:22Z

+        )
+    frid = match.group(1)
+
+    match = re.search(r"Module name:\s*(\S+)\n", commit_message)


nit: In this method you have duplicated module finding from commit message.

IMO it would be clearer if you would find the relevant commit first and then identify the module and have only one return.

pedjaradenkovic · 2026-05-13T07:01:46Z

+        return os.path.join(self.module_build_folder, CODEPLAIN_METADATA_FOLDER)
+
+    def get_module_render_status(self) -> tuple[str | None, str | None]:
+        if len(self.required_modules) == 0:


I think this is branch is not necessary and it's making confusion.

I'm not a big fun of recursions (Tjaz used to say that I'm old school) but in this case it can be written like:

module_name, frid = git_utils.get_last_rendered_functionality(self.module_build_folder) if module_name == self.module_name: return module_name, frid for required in reversed(self.required_modules): module_name, frid = required.get_module_render_status() if module_name is not None and frid is not None return module_name, frid return None, None

zanjonke requested a review from pedjaradenkovic April 21, 2026 09:17

zanjonke self-assigned this Apr 21, 2026

zanjonke added the enhancement New feature or request label Apr 21, 2026

zanjonke force-pushed the feat/partial_rendering branch 2 times, most recently from 7c4683a to b4929b2 Compare April 21, 2026 09:22

zanjonke force-pushed the feat/partial_rendering branch 10 times, most recently from 8e4bb3b to 9b096e6 Compare April 24, 2026 09:31

zanjonke force-pushed the feat/partial_rendering branch from 9b096e6 to 1700ddc Compare May 7, 2026 05:27

pedjaradenkovic reviewed May 7, 2026

View reviewed changes

pedjaradenkovic reviewed May 12, 2026

View reviewed changes

zanjonke force-pushed the feat/partial_rendering branch 5 times, most recently from 8770f97 to d4462c3 Compare May 15, 2026 12:58

zanjonke added 4 commits May 15, 2026 15:48

partial_rendering

473e500

Partial render choices and TUI

0b65edc

Adding tests

5e06ec8

Styling the PartialRenderTUI

ab39615

zanjonke added 2 commits May 15, 2026 15:55

Ask user when the only choice is destruction

d09d0d5

UI changes.

68346ec

zanjonke force-pushed the feat/partial_rendering branch from d4462c3 to 68346ec Compare May 15, 2026 13:55

pedjaradenkovic approved these changes May 15, 2026

View reviewed changes

zanjonke merged commit b9e0721 into main May 15, 2026
10 checks passed

Conversation

zanjonke commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Uh oh!

kaja-s commented Apr 22, 2026

Uh oh!

pedjaradenkovic left a comment

Choose a reason for hiding this comment

Uh oh!

kaja-s commented May 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pedjaradenkovic left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

zanjonke commented May 13, 2026

Uh oh!

pedjaradenkovic left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

zanjonke commented Apr 21, 2026 •

edited

Loading

kaja-s commented May 12, 2026 •

edited

Loading