+ + Coding Period Week 13: Aug 21 ~ Aug 26 + + +
+ + + + + +This is the final evaluation week, where ongoing and new errors were resolved, a demo video was created, and the summary report for the final evaluation of G...
+Goals
--
-
- Solve the Robotics Application with nested and composed version (Found Issues) -
Sub-Goals
--
-
- Understand David’s FSM circuit and issue -
- Replicate David’s VC Circuit with a newer version -
- Run the FSM on a single level -
- New issue: Nested parameters not appearing in the build file -
- Resolve the parameters issue by modifying earlier version code -
- New issue: Analyze the issue of nested FSM; unable to run single port with multiple IN and OUT. -
- Update the blog page with about, home, sidebar. -
Images
-Older version FSM single level circuit:
-
Recreated working FSM circuit with nested code:
-
-
Output of single level:
-
-
Parameter issue:
-
Single port - Multiple IN and Multiple OUT issue:
-
Accomplishment and Challenges
- --
-
-
-
Replicating David’s FSM Circuit
-The nested issue was opened by David around a year ago when he faced this problem while creating the FSM circuit with VC. My first goal was to recreate the nested FSM circuit. You may wonder why recreate it? The circuit was made with an earlier version, and the internal blocks will not have interdependency if I build with the older .vc3. So, I had to recreate every block and circuit. There was an issue when I ran the parameters; they were missing, so I checked with data.json and found missing parameters.
-
- -
-
Parameter Issue in Nested Code
-The earlier code was developed for one code block dependency, similar to nested blocks. The parameter IDs depended on the “type ID” of the dependency, but if multiple blocks are inside the dependency, this method will have the same type ID, and the parameters cannot map to the correct block. This type_ID needed to be changed in nested code. All blocks in data.json are at a single level, so the types are basic blocks. Initially, I thought to create a random alphanumeric string of 30 characters as the parameter ID. After implementing it, many issues were found, such as if parameters are used in input and output, leading to more confusion in wire mapping. Then, I implemented a new ID not dependent on type_ID, and particular wires will map with blocks, so the issue of mapping extra wires also won’t occur.
-
- -
-
Single level FSM Circuit
-After solving the parameters issue, I proceeded with the single-level FSM circuit. The blocks such as laser_ROS2, motor_ROS2, and screen used by David were not from the block library. Initially, it didn’t work, and finding the issue took more time. After analyzing the internal code of the above blocks, they were customized. After modifying and recreating, the whole logic ran without any issue. I ran turtlebot3 in an empty world inside my local machine and ran the FSM logic with VC.
-
- -
-
Issue in multi level FSM Circuit (Major issue)
-After running the single level, I was confident it would run with nested, but as the saying goes, things don’t always go as planned. David used single port multiple OUT to different blocks and took input outside of the composed block from multiple blocks. However, I wrote for single IN - OUT, so it could not map the multiple blocks; it was taking one block and connecting to another block. This is a big issue as it requires looking again at the mapping of wires and adding conditions for such situations.
-
-
PRs Created
- - -Issues
- - -Links
--
-
- Midterm update linkedin post - https://www.linkedin.com/feed/update/urn:li:activity:7226124206420180992/ -
Goals
--
-
- Multiple instances of block issue -
- Line follower with composed block feature -
- Partially resolved the single port with multiple input/output issue -
- Completed Python file for CI workflow build test -
Sub-Goals
--
-
- Solved the line follower with a single level -
- Resolved the single port with multiple input/output issue when the number of inputs and outputs is the same -
- Resolved the issue of identical IDs across multiple instances of the same block -
- Resolved the issue of model data remaining the same when the Back button is pressed -
- Solved the line follower with composed block feature -
- Completed the Python file for the CI workflow -
Images
-Composed circuit of line follower:
-
Internal perception composed block:
-
Global Input Output dialog box:
-
Accomplishment and Challenges
- --
-
-
-
Identical IDs across multiple instances of the same block
-In the line follower demo, Dr. Jose Maria suggested using two generic PID blocks, which led to the issue of identical IDs across multiple instances. While the line follower worked with a single PID, using two PID instances is more appropriate as it minimizes errors in linear and angular velocities. The issue arose because multiple levels were converted into a single level after backend processing, resulting in identical IDs for both internal blocks, causing incorrect wire and block mapping. To resolve this, I added an extra script in the “Add as Block” function before the jsonModel.design is appended. This script iterates to modify the IDs with random IDs, and not only the block but also the design’s wire and layer block IDs need to be changed, which was achieved in this extra script.
-
- -
-
Issue with model data remaining unchanged when the Back button is pressed
-After solving the issue with identical IDs across multiple instances, another issue arose where, upon entering a package, the ID and internal data of the block remained the same as they were at the time of importing. This issue was due to the node layers in the editor, as the model of the editor was appended in the higher-level block stack. The code I wrote to solve the multiple instance issue involved modifying the IDs of the design model. To fix this, I added an extra iteration in the loadPackage function. In the editor, there were two objects—one for links and one for blocks—so both IDs needed to be changed along with the earlier design IDs.
-
- -
-
Single port with multiple input/output issue
-This issue occurred when David’s FSM circuit was created using the composed block feature. In David’s circuit, multiple blocks were attached to a single port. My approach was to use a queue operation to map all blocks, but there were many cases where the number of inputs and outputs was not the same, requiring the creation of extra links. The issue with equal numbers of input/output blocks was resolved using the queue approach.
-
-
This GSoC project focuses on adding a global block composition feature, resolving nested issues, modifying the current version’s single dependency codebase to a multi-level dependency, implementing two or more demo robotics applications, expanding the block library, automating the testing of VisualCircuit and its documentation using GitHub Actions, and releasing the new version with all issue fixes.
The last 16 weeks have been an incredible and exhilarating journey filled with learning, challenges, and accomplishments. It was a great experience to work on such an exciting project, applying my web development skills to the field of Robotics, a domain I’m deeply passionate about, and learning more about both areas along the way. These past few months were exciting and memorable, with late nights, early mornings, and sometimes not sleeping at all. The joy of completing tasks is hard to describe in words (though my weekly blog attempts to capture some of that emotion).
@@ -316,183 +201,286 @@
Person Detector Block testing:
-
]]>Goals
-
-
- Create first global test workflow [In Progress] -
- Solve the issue of nested block [In Progress] +
- Develop generic PID and person detector blocks +
- Test new blocks by creating demo circuits +
- Refine the code and add comments +
- Resolve the issue with multiple instances of prebuilt package blocks +
- Create and edit demo video for final evaluation +
- Complete summary report and publish the video on youtube
Accomplishment and Challenges
-
-
Learning tools for Automation - Selenium
-To create a global test, I needed a tool that could automatically click respective options in the VisualCircuit application. I explored several technologies such as Puppeteer, Playwright, and Selenium. I chose Selenium as it is widely used for automation. I learned how Selenium works and how to use it.
-
- -
-
Integreation of Selenium with Github Action
-From my research, I observed that there are two ways to use Selenium with GitHub Actions: one involves the standard installation of Chrome and ChromeDriver, and the other uses the selenium/standalone-chrome Docker image. I chose the second method as it includes the latest Chrome and ChromeDriver, eliminating the need for a separate ChromeDriver. Additionally, it contains all the functionalities of the Chrome browser. Initially, the browser was not displaying correctly on the Selenium Grid, only showing the logo. To view the live action of the process, I opted for noVNC.
-
- -
-
Automating clicking File and Open buttons in the Menubar
-After successfully running selenium/standalone-chrome on my local machine, I automated the process of clicking the menubar buttons. I needed the XPATH to locate and perform the respective actions. I found the XPATH for the File and Open buttons from the Developer Console. Below is a GIF showing the automation process of opening the browser, launching VC, and clicking the respective buttons.
-
-
Generic PID and person detector block
+Now it’s time to add some blocks. First on the list is the Generic PID. A question that might arise is why we need this if there is already a PID block. Like you, I asked this question to my mentors, and they explained that the current PID block is designed for a particular use case where the PID is applied to angular velocity, while linear velocity remains constant. This PID block cannot be used for every case, which is why a generic PID block is needed that takes the error and provides a resulting value for both linear and angular velocity, not just angular. The generic PID can be used in any application.
-Automating clicking File and Open buttons in the Menubar:
+For the person detector block, I took reference from the Object Detector block. According to my mentor, we need a simple block that outputs an image with a selected box, indicating the detected person along with their coordinates, width, and height. Both blocks were tested by creating a demo application.
-
Generic PID and person detector block:
+
-
+
-
-
Research about Block Composition
-The first step was to understand the issue created by David Tapiador. I researched the data flow of opening prebuilt VC projects, adding as a block, and the build options. When the project is built, the dependency section includes package and design key-value pairs. The dependencies were not included, which may be the reason for this nested issue.
+Multiple Instances Issue with Prebuilt Package Block
+This unexpected error occurred in the code for handling multiple instances from the previous week when blocks were added from the prebuilt collection block. Upon verification, it was observed that the code I wrote was correct and should work for every package block since it was running in loadPackage, which is also called in the collection block. Therefore, the issue is not with the code itself, but rather with the debugging process. It was observed that when one collection block is added, the ID changes, but when adding another identical collection block, the ID changes again, altering the previous block ID. This seems to be modifying the entire model, and the reason is still unclear but needs to be investigated further.
-
-
Adding interdependencies
-For nested blocks, I modified the Dependency interface to include the interdependencies element. I also modified the package-model and converter for dependencies. Successfully, interdependencies appeared inside dependencies, but they were blank. This blank state is a great checkpoint as modifying internal structures introduced many errors. I am happy with the progress of this blank interdependencies issue.
+Demo video and summary report
+Almost everything was aligned correctly for the demo video, with some loose ends that are extra features and minor unexpected errors that can be resolved in the future. Apart from that, everything is functioning as expected. It was time to create the demo video showcasing the features that will be released in v3.6. In the video, I demonstrated the state before v3.6, the modifications made in v3.6, and how to use the composed and nested block features. After verification from my mentors, I published the video on YouTube and also wrote the summary report required for the final evaluation of GSoC.
+ + +
Inside person detector block:
+
Integreation of interdependencies inside dependencies:
+Final Note
+With this, my GSoC work is complete, but I believe this is just the beginning of a new chapter with JdeRobot and open source. I will continue contributing as time allows, alongside my other commitments.
-
-
If you’ve been following from the beginning, thank you for reading up to week 13—I truly appreciate your time. If you started with this blog or any intermediate blog, you can read all the posts from day 0 to the present. A summary blog is also available, which covers the entire GSoC project progress. If you have any questions or comments, feel free to reach out to me through my social media handles provided in the sidebar. Again, thank you for visiting my blog and reading about my GSoC journey.
]]>Commits
--
-
- https://github.com/JdeRobot/VisualCircuit/commit/7ccbea013a311fe1d0af6e71408b84b3b7594f15 -
- https://github.com/JdeRobot/VisualCircuit/commit/8ea1c0d35e1ee11f282a7407635ff899c538f34f -
Previous PR Updates
--
-
- Merged - https://github.com/JdeRobot/VisualCircuit/pull/309 -
Goals
-
-
- Modification of VC block compostion for Global Input Output [In Progress] -
- Create first global test workflow [In Progress] -
- Solve the issue of nested block [In Progress] +
- Multiple instances of block issue +
- Line follower with composed block feature +
- Partially resolved the single port with multiple input/output issue +
- Completed Python file for CI workflow build test +
Sub-Goals
+-
+
- Solved the line follower with a single level +
- Resolved the single port with multiple input/output issue when the number of inputs and outputs is the same +
- Resolved the issue of identical IDs across multiple instances of the same block +
- Resolved the issue of model data remaining the same when the Back button is pressed +
- Solved the line follower with composed block feature +
- Completed the Python file for the CI workflow
Images
+Composed circuit of line follower:
+
Internal perception composed block:
+
Global Input Output dialog box:
+
Accomplishment and Challenges
-
-
Understanding the VC Block Composition
-I went through some frontend files such as
+menu/index.tsx,project-info-dialog.tsx, andeditor.ts. These files are responsible for editing the project information on the frontend side. Initially, I thought of adding Global Input Output functionality within the existing project editing interface. However, as I began implementing it, I realized that it was more difficult than anticipated since it required changing the Project Info interface in every file. Therefore, I decided to build a new dialog box that can be opened from the “Save Block” option in the menu bar.Identical IDs across multiple instances of the same block
+In the line follower demo, Dr. Jose Maria suggested using two generic PID blocks, which led to the issue of identical IDs across multiple instances. While the line follower worked with a single PID, using two PID instances is more appropriate as it minimizes errors in linear and angular velocities. The issue arose because multiple levels were converted into a single level after backend processing, resulting in identical IDs for both internal blocks, causing incorrect wire and block mapping. To resolve this, I added an extra script in the “Add as Block” function before the jsonModel.design is appended. This script iterates to modify the IDs with random IDs, and not only the block but also the design’s wire and layer block IDs need to be changed, which was achieved in this extra script.
-
-
Implementation of Global Input Output on frontend
-For choosing the Global Input Output, all inputs and outputs can be shown, and from there, the Global Input Output can be assigned. I created a new script for the dialog box and modified
+editor.tsandmenu/index.tsx. Currently, all inputs and outputs are displayed. Next week, I will work on assigning the inputs and outputs as Global for the block.Issue with model data remaining unchanged when the Back button is pressed
+After solving the issue with identical IDs across multiple instances, another issue arose where, upon entering a package, the ID and internal data of the block remained the same as they were at the time of importing. This issue was due to the node layers in the editor, as the model of the editor was appended in the higher-level block stack. The code I wrote to solve the multiple instance issue involved modifying the IDs of the design model. To fix this, I added an extra iteration in the loadPackage function. In the editor, there were two objects—one for links and one for blocks—so both IDs needed to be changed along with the earlier design IDs.
+ -
+
Single port with multiple input/output issue
+This issue occurred when David’s FSM circuit was created using the composed block feature. In David’s circuit, multiple blocks were attached to a single port. My approach was to use a queue operation to map all blocks, but there were many cases where the number of inputs and outputs was not the same, requiring the creation of extra links. The issue with equal numbers of input/output blocks was resolved using the queue approach.
+
+
Goals
+-
+
- Solve the Robotics Application with nested and composed version (Found Issues)
Checkbox for global input output - dialog box:
+Sub-Goals
+-
+
- Understand David’s FSM circuit and issue +
- Replicate David’s VC Circuit with a newer version +
- Run the FSM on a single level +
- New issue: Nested parameters not appearing in the build file +
- Resolve the parameters issue by modifying earlier version code +
- New issue: Analyze the issue of nested FSM; unable to run single port with multiple IN and OUT. +
- Update the blog page with about, home, sidebar. +
Images
+Older version FSM single level circuit:
+
Recreated working FSM circuit with nested code:
+
+
Output of single level:
+
+
Parameter issue:
+
Single port - Multiple IN and Multiple OUT issue:
+
Accomplishment and Challenges
-
-
Solving the nested block with different approach
-In this approach, I didn’t use
+node.dependenciesbut instead created blocks and wires for internal dependencies. This solution was suggested by ChatGPT, so I tried it. It involved creating internal wires for dependencies and making the linkages. I spent around two days on this, but it didn’t yield the desired results. Additionally, it was suggesting editing many internal components, so I dropped this idea and returned to my nested block solution.Replicating David’s FSM Circuit
+The nested issue was opened by David around a year ago when he faced this problem while creating the FSM circuit with VC. My first goal was to recreate the nested FSM circuit. You may wonder why recreate it? The circuit was made with an earlier version, and the internal blocks will not have interdependency if I build with the older .vc3. So, I had to recreate every block and circuit. There was an issue when I ran the parameters; they were missing, so I checked with data.json and found missing parameters.
-
-
Solving the nested block with last week approach
-In the previous week, I encountered blank dependencies. I noticed that the interface dependencies contained a single dependency, so I modified it to include an array and used recursive calls to add internal dependencies. Additionally,
+factory.tsxhad not been changed, so I modified that as well. However, the issue of blank dependencies still persists. I need to consult the mentors regarding this issue, as I feel it can be resolved with some more effort.Parameter Issue in Nested Code
+The earlier code was developed for one code block dependency, similar to nested blocks. The parameter IDs depended on the “type ID” of the dependency, but if multiple blocks are inside the dependency, this method will have the same type ID, and the parameters cannot map to the correct block. This type_ID needed to be changed in nested code. All blocks in data.json are at a single level, so the types are basic blocks. Initially, I thought to create a random alphanumeric string of 30 characters as the parameter ID. After implementing it, many issues were found, such as if parameters are used in input and output, leading to more confusion in wire mapping. Then, I implemented a new ID not dependent on type_ID, and particular wires will map with blocks, so the issue of mapping extra wires also won’t occur.
+
+ -
+
Single level FSM Circuit
+After solving the parameters issue, I proceeded with the single-level FSM circuit. The blocks such as laser_ROS2, motor_ROS2, and screen used by David were not from the block library. Initially, it didn’t work, and finding the issue took more time. After analyzing the internal code of the above blocks, they were customized. After modifying and recreating, the whole logic ran without any issue. I ran turtlebot3 in an empty world inside my local machine and ran the FSM logic with VC.
+
+ -
+
Issue in multi level FSM Circuit (Major issue)
+After running the single level, I was confident it would run with nested, but as the saying goes, things don’t always go as planned. David used single port multiple OUT to different blocks and took input outside of the composed block from multiple blocks. However, I wrote for single IN - OUT, so it could not map the multiple blocks; it was taking one block and connecting to another block. This is a big issue as it requires looking again at the mapping of wires and adding conditions for such situations.
Previous PR Updates
+PRs Created
]]>Issues
+ + +Links
+-
+
- Midterm update linkedin post - https://www.linkedin.com/feed/update/urn:li:activity:7226124206420180992/ +
Goals
-
-
- Complete the first Selenium-based global test -
- Modification of VC block compostion for Global Input Output +
- Solve the nested issue completely +
- Integrate nested code into last week composed block code. +
Sub-Goals
+-
+
- Understanding backend processing for single-level dependencies. +
- Creating sample tests to determine necessary wires and blocks for the built zip file. +
- Fixing the issue of missing nested wires in data.json after backend processing. +
- Addressing the problem of missing nested internal blocks in data.json. +
- Modifying the current version’s single code block dependency code to support multi-level dependencies with multiple code blocks. +
- Writing code for processing wires, including filtering and mapping to required blocks. +
- Removing unwanted links between global inputs outputs and testing the nested issue code with different examples.
Images
+Level1 Sample Circuit:
+
Level2 Inside test_second_block Block (nested level 1):
+
Level3 Inside test_first_block Block (nested level 2):
+
Before modification of nested issue - data.json:
+
After modification of nested issue - data.json:
+
Python Modules Scripts after processing:
+
Output including printing at every level and code block:
+
Accomplishment and Challenges
-
-
Pasting .vc3 file and opening the file in VC
-I first tried putting the circuit file in the Selenium Docker container through a Python script, but it wasn’t a great solution, so I decided to put the circuit file through YAML using Docker commands. After solving this, I needed to open the file from a particular directory in the file manager dialog box. For this, OS-level control was needed, so I used the pyautogui library for opening the file in Visual Circuit.
-
- -
-
Building and solving inseure download issue
-There were many errors when building the project as the backend was running on localhost on the local machine and had to be accessed from the Docker container, resulting in fetch errors. These were resolved after modifying setting.py and .env. After that, the browser wasn’t allowing downloads as Chrome considered the HTTP frontend insecure. After some Googling and searching on StackOverflow, I found a solution. Finally, the whole process was completed automatically by Selenium. The complete process is shown in the GIF below. -Selenium based first automated global test:
+Modifying the Current Version to Support Multi Code Block Dependency
When I completed writing the entire logic, I found that multiple codes were not supported, and extra wires were created that couldn’t connect. Additionally, other code blocks of inter-dependency were not included in data.json, leading to improper linkage. I modified synthesize_modules and changed the logic for adding internal blocks, so they were not treated as whole dependencies. This way, the blocks were added at one level.
-
-
Improving fetch Global Input and Output function
-In the previous week, I was able to fetch the port data from all blocks, but it was using more iterations and not proper internal functions for fetching the data. So I improved the global input/output function and also added the submit and checklist handlers to pass data to other scripts.
-
- -
-
Creating input output blocks automatically from checkbox data
-Depending on the port chosen by the user, the particular global input/output block is generated automatically. It checks the port type, and according to that, the input or output block is generated with the same name. Initially, the global input/output block position is the same as the initial position, and the links are null.
+Understanding Backend Processing by Creating Sample Tests
+To understand the process, I created a sample test and added blocks and wires in data.json. This helped me see how the internal file structure of the zip file works after building. I drew the whole connection and tried to replicate it by modifying synthesis.py using recursive and iterative techniques until level 1 was achieved.
-
-
Modification of Links and creating respective input output block
-I researched the internal structure of the model and tried to access the links and respective link IDs of each block. After this, I wrote down all possible cases for creating a circuit by the user. One case is where the global input/output has zero links, so one link has to be created, and the source and target assigned. Another case is that there is already a link connected with the block, so according to the port type, the particular source or target has to be removed and a new one assigned. Apart from that, there are other cases that need to be tested in the future, such as if there are two links at the same port, which one to choose. I want to ask the mentors about these cases.
+Fixing the Issue of Missing Nested Blocks After Backend Processing
+I created the process_dependency function, which is called inside synthesize_modules. This function iterates over all blocks of a dependency and creates module Python scripts depending on the block type. Only code block type modules and block values are included in data.json. If there is a dependency inside a dependency, process_dependency is called recursively until no nested dependencies are left. This method ensures that only code blocks are included in the modules directory, while other blocks such as global input/output, dependencies, and constants are filtered out.
-
-
Saving and retriving old blocks function
-Before generating the .vc3 composed block file, the old model has to be saved so the user can continue working on the circuit without worrying about the blocks global input/output. The model and project info are saved in the stack, and whenever the build is completed, the old model is loaded. There were constant errors, such as storing the data in the stack but it was changing with changes, as I was storing this.activModel directly in the stack, causing changes in the stack. For that, I tried several methods using deepCopy, custom functions, and cloneDeep. Both deepCopy and custom functions didn’t work, so I used the cloneDeep library, which worked without any errors. Similar to the “Save as” functionality, I used that part of the code for generating .vc3 file.
+Writing code for filtering and mapping the wires in backend
+After obtaining the proper blocks in data.json, I focused on the wires. In the current version, level 1 blocks were being added to data.json, so I modified process_dependency to include every wire. There were wires for packages and global input/output, which required proper filtering and the creation of new links between the required blocks. I iterated over all wires selected from process_dependency and segregated them based on the wire’s source and target block ID. After that, in another iteration, I checked each wire with the filtered blocks in data.json to determine if the block was present, adding the key pair of ob:absent. Then, I created process_wire, where all wire mappings were handled. This function contains nested iterations:
Normal circuit to be converted for block:
-
Global input ouput port selection :
-
-
+
- The first internal iteration iterates over all valid wires, checks for global input and output, and either saves the wire in a dictionary or removes the entire source-target link based on the presence of the absent value. +
- The second internal iteration checks the wire’s port, replacing package-type wire with the corresponding source-target dictionary generated in the last iteration. +
Generated block:
-
This way, new links are created, and the absent value helps the code decide whether to loop again or exit process_wire.
-Internal block structure:
-
PRs Created
+-
+
- Closed and Merged - https://github.com/JdeRobot/VisualCircuit/pull/322 +
- Closed and Merged - https://github.com/JdeRobot/VisualCircuit/pull/323 +
Commits
]]>Goals
-
-
- Modify the composed block to follow specific rules -
- Refine the previous block composition code -
- Create demo applications with composed blocks +
- Solve the blank dependencies issue - nested issue +
- Resolve the issue of nested Python files not appearing in the Modules directories - nested issue
Images
+Nested dependencies in .vc3/ .json file Level 2:
+
Nested dependencies in .vc3/ .json file Level 3:
+
Modules directory with all nested python files:
+
Accomplishment and Challenges
-
-
Solving the global block input links issue and extending block composition for every block
-As discussed above, we should not have more than one link at the input port, and we need to hide input ports with internal wiring from the block composition dialog box. I modified the globalInputOutput function to segregate ports based on links and port types. In last week’s code, only code blocks were supported, so I made changes to support every block. I also observed an issue with the text being fetched from the package, which was taking a 40-character name. I modified it to take the name port from the labels of the blocks.
+Analyzing the Week 3 Code for the Nested Issue
+Dr. Jose Maria suggested in the meeting to revisit the basics and look for solutions in the internal architecture. Therefore, I recreated the issue from week 3. I wrote two codes: one where a new node is created according to dependencies and new wiring was made. However, this approach was not sustainable due to the potential complexity from multiple dependencies. I dropped that idea. The second approach was recursion-based on frontend side, aimed at getting all dependencies, but it resulted in blank dependencies. After analyzing this solution, I changed the recursion solution from the package level.
-
-
Solving the global block output links issue and refining existing code
-For blocks with multiple wiring at the output side, a new link needs to be added with the respective global input/output attached. Unlike the global input block issue, this requires the parent global output block ID and the respective link ID so that the existing link is not modified. These changes were extensive and altered last week’s code, so I refined the overall code. After refinement, the entire global input/output block was tested with all cases and is working as expected.
+Blank dependicies issue
+Upon deeper analyzing into PackageBlockModel and factory.tsx, I observed that I had created nested dependencies that were not present in the interface of PackageBlockModel. I modified the dependency array for every script and added internal dependencies to the respective interface and methods. The solution was close to being resolve, but I took a different approach in week 3.
+
+ -
+
Python script in modules directory issue
+Initially, the circuit was not building, but the Save as function was working. Due to nested dependencies, the synthesis file could not be processed. I added a recursive solution to get the nested dependencies from the frontend, which was very challenging. The files were being created but not at all levels. After debugging, I found that the issue was the missing internal dependencies at each level and the improper assignment of keys. The above images attached show the progress made this week.
Commits
+PR Created
]]>Goals
- Develop the Robotics Application with composed block @@ -529,125 +517,126 @@ Selenium based first automated global test:
Blank script and no internal script in the modules directory issue image:
-
]]>Goals
-
-
- Solve the blank dependencies issue - nested issue -
- Resolve the issue of nested Python files not appearing in the Modules directories - nested issue +
- Modify the composed block to follow specific rules +
- Refine the previous block composition code +
- Create demo applications with composed blocks
Images
-Nested dependencies in .vc3/ .json file Level 2:
-
Nested dependencies in .vc3/ .json file Level 3:
-
Modules directory with all nested python files:
-
Accomplishment and Challenges
-
-
Analyzing the Week 3 Code for the Nested Issue
-Dr. Jose Maria suggested in the meeting to revisit the basics and look for solutions in the internal architecture. Therefore, I recreated the issue from week 3. I wrote two codes: one where a new node is created according to dependencies and new wiring was made. However, this approach was not sustainable due to the potential complexity from multiple dependencies. I dropped that idea. The second approach was recursion-based on frontend side, aimed at getting all dependencies, but it resulted in blank dependencies. After analyzing this solution, I changed the recursion solution from the package level.
-
- -
-
Blank dependicies issue
-Upon deeper analyzing into PackageBlockModel and factory.tsx, I observed that I had created nested dependencies that were not present in the interface of PackageBlockModel. I modified the dependency array for every script and added internal dependencies to the respective interface and methods. The solution was close to being resolve, but I took a different approach in week 3.
+Solving the global block input links issue and extending block composition for every block
+As discussed above, we should not have more than one link at the input port, and we need to hide input ports with internal wiring from the block composition dialog box. I modified the globalInputOutput function to segregate ports based on links and port types. In last week’s code, only code blocks were supported, so I made changes to support every block. I also observed an issue with the text being fetched from the package, which was taking a 40-character name. I modified it to take the name port from the labels of the blocks.
-
-
Python script in modules directory issue
-Initially, the circuit was not building, but the Save as function was working. Due to nested dependencies, the synthesis file could not be processed. I added a recursive solution to get the nested dependencies from the frontend, which was very challenging. The files were being created but not at all levels. After debugging, I found that the issue was the missing internal dependencies at each level and the improper assignment of keys. The above images attached show the progress made this week.
+Solving the global block output links issue and refining existing code
+For blocks with multiple wiring at the output side, a new link needs to be added with the respective global input/output attached. Unlike the global input block issue, this requires the parent global output block ID and the respective link ID so that the existing link is not modified. These changes were extensive and altered last week’s code, so I refined the overall code. After refinement, the entire global input/output block was tested with all cases and is working as expected.
PR Created
+Commits
]]>Goals
-
-
- Solve the nested issue completely -
- Integrate nested code into last week composed block code. +
- Complete the first Selenium-based global test +
- Modification of VC block compostion for Global Input Output
Sub-Goals
--
-
- Understanding backend processing for single-level dependencies. -
- Creating sample tests to determine necessary wires and blocks for the built zip file. -
- Fixing the issue of missing nested wires in data.json after backend processing. -
- Addressing the problem of missing nested internal blocks in data.json. -
- Modifying the current version’s single code block dependency code to support multi-level dependencies with multiple code blocks. -
- Writing code for processing wires, including filtering and mapping to required blocks. -
- Removing unwanted links between global inputs outputs and testing the nested issue code with different examples. +
-
+
Pasting .vc3 file and opening the file in VC
+I first tried putting the circuit file in the Selenium Docker container through a Python script, but it wasn’t a great solution, so I decided to put the circuit file through YAML using Docker commands. After solving this, I needed to open the file from a particular directory in the file manager dialog box. For this, OS-level control was needed, so I used the pyautogui library for opening the file in Visual Circuit.
+
+ -
+
Building and solving inseure download issue
+There were many errors when building the project as the backend was running on localhost on the local machine and had to be accessed from the Docker container, resulting in fetch errors. These were resolved after modifying setting.py and .env. After that, the browser wasn’t allowing downloads as Chrome considered the HTTP frontend insecure. After some Googling and searching on StackOverflow, I found a solution. Finally, the whole process was completed automatically by Selenium. The complete process is shown in the GIF below. +Selenium based first automated global test:
+ -
+
Improving fetch Global Input and Output function
+In the previous week, I was able to fetch the port data from all blocks, but it was using more iterations and not proper internal functions for fetching the data. So I improved the global input/output function and also added the submit and checklist handlers to pass data to other scripts.
+
+ -
+
Creating input output blocks automatically from checkbox data
+Depending on the port chosen by the user, the particular global input/output block is generated automatically. It checks the port type, and according to that, the input or output block is generated with the same name. Initially, the global input/output block position is the same as the initial position, and the links are null.
+
+ -
+
Modification of Links and creating respective input output block
+I researched the internal structure of the model and tried to access the links and respective link IDs of each block. After this, I wrote down all possible cases for creating a circuit by the user. One case is where the global input/output has zero links, so one link has to be created, and the source and target assigned. Another case is that there is already a link connected with the block, so according to the port type, the particular source or target has to be removed and a new one assigned. Apart from that, there are other cases that need to be tested in the future, such as if there are two links at the same port, which one to choose. I want to ask the mentors about these cases.
+
+ -
+
Saving and retriving old blocks function
+Before generating the .vc3 composed block file, the old model has to be saved so the user can continue working on the circuit without worrying about the blocks global input/output. The model and project info are saved in the stack, and whenever the build is completed, the old model is loaded. There were constant errors, such as storing the data in the stack but it was changing with changes, as I was storing this.activModel directly in the stack, causing changes in the stack. For that, I tried several methods using deepCopy, custom functions, and cloneDeep. Both deepCopy and custom functions didn’t work, so I used the cloneDeep library, which worked without any errors. Similar to the “Save as” functionality, I used that part of the code for generating .vc3 file.
+
+
Accomplishment and Challenges
+ +-
+
Images
-Level1 Sample Circuit:
-
Level2 Inside test_second_block Block (nested level 1):
-
-
+
Level3 Inside test_first_block Block (nested level 2):
-
Normal circuit to be converted for block:
+
Before modification of nested issue - data.json:
-
Global input ouput port selection :
+
After modification of nested issue - data.json:
-
Generated block:
+
Python Modules Scripts after processing:
-
Internal block structure:
+
Output including printing at every level and code block:
-
Commits
+]]>Goals
+-
+
- Modification of VC block compostion for Global Input Output [In Progress] +
- Create first global test workflow [In Progress] +
- Solve the issue of nested block [In Progress] +
Accomplishment and Challenges
-
-
Modifying the Current Version to Support Multi Code Block Dependency
+Understanding the VC Block Composition
+I went through some frontend files such as
+menu/index.tsx,project-info-dialog.tsx, andeditor.ts. These files are responsible for editing the project information on the frontend side. Initially, I thought of adding Global Input Output functionality within the existing project editing interface. However, as I began implementing it, I realized that it was more difficult than anticipated since it required changing the Project Info interface in every file. Therefore, I decided to build a new dialog box that can be opened from the “Save Block” option in the menu bar.
+ -
+
Implementation of Global Input Output on frontend
+For choosing the Global Input Output, all inputs and outputs can be shown, and from there, the Global Input Output can be assigned. I created a new script for the dialog box and modified
editor.tsandmenu/index.tsx. Currently, all inputs and outputs are displayed. Next week, I will work on assigning the inputs and outputs as Global for the block.
When I completed writing the entire logic, I found that multiple codes were not supported, and extra wires were created that couldn’t connect. Additionally, other code blocks of inter-dependency were not included in data.json, leading to improper linkage. I modified synthesize_modules and changed the logic for adding internal blocks, so they were not treated as whole dependencies. This way, the blocks were added at one level.
+Checkbox for global input output - dialog box:
+ +
-
-
Understanding Backend Processing by Creating Sample Tests
-To understand the process, I created a sample test and added blocks and wires in data.json. This helped me see how the internal file structure of the zip file works after building. I drew the whole connection and tried to replicate it by modifying synthesis.py using recursive and iterative techniques until level 1 was achieved.
-
- -
-
Fixing the Issue of Missing Nested Blocks After Backend Processing
-I created the process_dependency function, which is called inside synthesize_modules. This function iterates over all blocks of a dependency and creates module Python scripts depending on the block type. Only code block type modules and block values are included in data.json. If there is a dependency inside a dependency, process_dependency is called recursively until no nested dependencies are left. This method ensures that only code blocks are included in the modules directory, while other blocks such as global input/output, dependencies, and constants are filtered out.
+Solving the nested block with different approach
+In this approach, I didn’t use
node.dependenciesbut instead created blocks and wires for internal dependencies. This solution was suggested by ChatGPT, so I tried it. It involved creating internal wires for dependencies and making the linkages. I spent around two days on this, but it didn’t yield the desired results. Additionally, it was suggesting editing many internal components, so I dropped this idea and returned to my nested block solution. -
-
Writing code for filtering and mapping the wires in backend
-After obtaining the proper blocks in data.json, I focused on the wires. In the current version, level 1 blocks were being added to data.json, so I modified process_dependency to include every wire. There were wires for packages and global input/output, which required proper filtering and the creation of new links between the required blocks. I iterated over all wires selected from process_dependency and segregated them based on the wire’s source and target block ID. After that, in another iteration, I checked each wire with the filtered blocks in data.json to determine if the block was present, adding the key pair of ob:absent. Then, I created process_wire, where all wire mappings were handled. This function contains nested iterations:
+Solving the nested block with last week approach
+In the previous week, I encountered blank dependencies. I noticed that the interface dependencies contained a single dependency, so I modified it to include an array and used recursive calls to add internal dependencies. Additionally,
factory.tsxhad not been changed, so I modified that as well. However, the issue of blank dependencies still persists. I need to consult the mentors regarding this issue, as I feel it can be resolved with some more effort.
Previous PR Updates
-
-
- The first internal iteration iterates over all valid wires, checks for global input and output, and either saves the wire in a dictionary or removes the entire source-target link based on the presence of the absent value. -
- The second internal iteration checks the wire’s port, replacing package-type wire with the corresponding source-target dictionary generated in the last iteration. -
This way, new links are created, and the absent value helps the code decide whether to loop again or exit process_wire.
- -PRs Created
--
-
- Closed and Merged - https://github.com/JdeRobot/VisualCircuit/pull/322 -
- Closed and Merged - https://github.com/JdeRobot/VisualCircuit/pull/323 -