Hi Kirk,
thank you for your investigation and clarification. And thanks to Claude for his nice overview. Stil then the most valuable input for me was to use another command (without knowing exactly what this means for the company restrictions):
py -m pip install beautifulsoup4
this worked with the same settings, i.e. without any additional admin rights or so.
Great! Glad it worked. I hope the scripts work well for you. 
P.S. I see now that you had already written that. Sorry to have missed that.
What does py -m pip install beautifulsoup4 do?
The command
py -m pip install beautifulsoup4 installs the Beautiful Soup 4 library, a tool used in Python for web scraping and parsing HTML and XML documents.
Here’s a breakdown of the command:
py: On Windows, this runs the Python launcher and executes the command for the version of Python you have installed. On other operating systems, you would typically use python3 instead.-m: This flag tells Python to run the following module as a script.pip: The package installer for Python, which manages and installs libraries not included in the standard Python installation.install: The specific pip command to download and install a package.beautifulsoup4: The name of the package to be installed from the Python Package Index (PyPI). The library is then imported in your code from the bs4 module.What Beautiful Soup is used for
Example usage
After installing, you can import and use the library in a Python script to parse HTML:
python
from bs4 import BeautifulSoup
import requests
# Fetch the HTML content from a website
response = requests.get("https://news.ycombinator.com/")
html_content = response.content
# Create a BeautifulSoup object to parse the HTML
soup = BeautifulSoup(html_content, "html.parser")
# Find the first title tag
title_tag = soup.find("title")
# Print the text content of the title tag
print(title_tag.text)
Use code with caution.
I recently got a question on GitHub from someone who didn’t know where to find the HTML documentation for an app. I have a text explanation here
but I neglected to provide a screenshot. Fortunately, @Steve has already provided a clear visual so I’ll share that here:
Analyze your CSV files in Google SheetsIf you’d like to use Google Sheets functions such as QUERY(), FILTER(), or REGEXMATCH() to search and analyze the CSV files produced by the AppSheet Parser Suite, this script can save you a lot of manual work.
Why It’s UsefulAppSheet’s HTML export, once parsed by the Parser Suite, produces several CSVs (appsheet_columns.csv, appsheet_views.csv, etc.).
Analyzing relationships or searching across them inside Google Sheets can be powerful, but importing each file by hand is tedious and error-prone.
This Google Apps Script automatically creates a Google Sheets file containing the parsed data from your app. Expressions are brought into the file as text to prevent errors and ensure that they will be searchable. Once converted to a Google Sheets file, you can use spreadsheet tools (QUERY, FILTER, BYROW, etc.) to explore your app.
PreparationBefore running the script, upload your parsed CSV files to Google Drive:
https://drive.google.com/drive/folders/1A2b3C4D5E6F7G8H9I0J
const folderInput = "PASTE_YOUR_FOLDER_URL_OR_ID_HERE";
Script// --- begin script ---
/**
* Import all CSV files from a Google Drive folder into one spreadsheet.
* - Spreadsheet name = folder name (exactly)
* - Sheet names = CSV file names (without .csv)
* - All values imported as literal text
* - Result link shown in Execution Log
*/
function importCSVsFromFolderDirect() {
const folderInput = "PASTE_YOUR_FOLDER_URL_OR_ID_HERE";
const folderId = extractFolderId(folderInput);
if (!folderId) {
Logger.log("❌ Invalid folder URL or ID. Please check the line near the top of the script.");
return;
}
const link = importCSVsFromFolder(folderId);
Logger.log("✅ Done! Created spreadsheet: " + link);
}
function importCSVsFromFolder(folderId) {
const folder = DriveApp.getFolderById(folderId);
const folderName = folder.getName();
const ss = SpreadsheetApp.create(folderName);
const files = folder.getFiles();
let csvCount = 0;
while (files.hasNext()) {
const file = files.next();
const name = file.getName();
if (!/\.csv$/i.test(name)) continue;
const text = stripBOM(file.getBlob().getDataAsString('UTF-8'));
const delimiter = detectDelimiter(text);
const rows = Utilities.parseCsv(text, delimiter) || [];
// Treat all cells as literal text
const safeRows = rows.map(row =>
row.map(cell => {
let value = String(cell);
if (/^[=+]/.test(value)) value = "'" + value; // escape formulas
return value;
})
);
const sheetName = makeUniqueSheetName(ss, sanitizeSheetName(name.replace(/\.csv$/i, '')));
const sheet = ss.insertSheet(sheetName);
const range = sheet.getRange(1, 1, Math.max(1, safeRows.length), Math.max(1, safeRows[0]?.length || 1));
range.setNumberFormat('@');
if (safeRows.length > 0 && safeRows[0].length > 0) {
range.setValues(safeRows);
}
csvCount++;
}
const defaultSheet = ss.getSheets()[0];
if (defaultSheet.getDataRange().getValue() === '') ss.deleteSheet(defaultSheet);
if (csvCount === 0) Logger.log("⚠️ No CSV files found in this folder.");
return ss.getUrl();
}
/** ---------- Helpers ---------- **/
function extractFolderId(input) {
const urlMatch = input.match(/\/folders\/([a-zA-Z0-9_-]+)/);
if (urlMatch) return urlMatch[1];
const qMatch = input.match(/[?&]id=([a-zA-Z0-9_-]+)/);
if (qMatch) return qMatch[1];
if (/^[a-zA-Z0-9_-]{20,}$/.test(input)) return input;
return null;
}
function stripBOM(text) {
return text.charCodeAt(0) === 0xFEFF ? text.slice(1) : text;
}
function detectDelimiter(text) {
const firstLine = (text.split(/\r?\n/).find(l => l.trim().length) || '');
const counts = { ',': (firstLine.match(/,/g) || []).length, '\t': (firstLine.match(/\t/g) || []).length, ';': (firstLine.match(/;/g) || []).length };
let best = ',', max = -1;
for (const d in counts) if (counts[d] > max) { max = counts[d]; best = d; }
return best;
}
function sanitizeSheetName(name) {
let n = name.replace(/[\[\]\:\*\?\/\\]/g, ' ');
if (n.length > 100) n = n.slice(0, 100);
return n.trim() || 'Sheet';
}
function makeUniqueSheetName(ss, base) {
let name = base, i = 2;
while (ss.getSheetByName(name)) {
const suffix = ' (' + i + ')';
const head = base.length + suffix.length > 100 ? base.slice(0, 100 - suffix.length) : base;
name = head + suffix;
i++;
}
return name;
}
// --- end script ---
How to Use
Save, then give the project a name such as CSV Folder Importer.const folderInput = "PASTE_YOUR_FOLDER_URL_OR_ID_HERE";
https://drive.google.com/drive/folders/1A2b3C4D5E6F7G8H9I0JimportCSVsFromFolderDirect, then click the 
Run button.
✅ Done! Created spreadsheet: https://docs.google.com/spreadsheets/d/xxxxxxxxxxxx/edit
Next StepsOnce you have the combined Google Sheet, you can use spreadsheet formulas to explore your AppSheet data.
Tip: Once you’ve verified that everything imports correctly, you can reuse this script anytime—just change the folder URL near the top and run it again to generate a fresh Google Sheet for new projects or new versions of the same project.
P.S. To save time, I used ChatGPT to prepare this post and the script. I have confirmed that the script works with my own data and I have read and revised the text of this post to my satisfaction. Any remaining problems are my fault, of course, not the responsibility of any AI tool I used.
You are really addressing a need here with this app/scripts. Thanks @Kirk_Masden !
At my first attempt today, everything seemed good up to Phase 4, where the parser failed!?
Any idea on how I can fix this would be greatly appreciated.
PHASE 4: Parsing Actions
Loading slice mapping from appsheet_slices.csv
Loaded 27 slice mappings
Extracting actions…
Loading system action data from actions.txt
Found 343 actions in text file (102 system-generated)
Actions parser failed: sequence item 0: expected str instance, NoneType found
Parsing complete!
Summary of data to be analyzed:
Data Sources: 60 total
├─ Tables: 33
└─ Slices: 27
Columns: 616 total
├─ Physical: 569
└─ Virtual: 47
Format rules: 37
Actions: Failed
Error: sequence item 0: expected str instance, NoneType found
Views: 122 total
├─ System-generated: 72
└─ User-created: 50
Navigation targets: 0
hi @shimodabt !
Thank you for using the scripts. Apologies for the error you encountered. I think the script is probably not robust enough to deal with your particular configuration.
I wonder if you would be willing to share your data with me so that I can try to parse it and investigate the problem.
Unless I’m mistaken, private messaging in not enabled on this forum so we need to contact each other via email. kirkmasden@hotmail.com is not my regular email account (I rarely look at it) but I’ll try to check it over the next week. If you contact me there, we can discuss how to share data confidentially.
Email sent @Kirk_Masden
Thanks! I sent a reply and Cc’d the Gmail account I would like to use to collaborate on this. I’m looking forward to finding the problem and improving the scripts. 
Gmail put my message in the “Spam” folder. So, you may need to look there to find it.
I found the source of the problem!
I was able to use the data you kindly shared with me to track down exactly what was happening. The issue occurs when an app has an action configured as “Execute an action on a set of rows” but the target action hasn’t been specified yet — in the JSON, this shows up as "ReferencedAction":null. The parser wasn’t handling that null value properly, which caused the crash.
If you want to fix this yourself right away, open actions_parser.py and find line 708. Change:
if 'ReferencedAction' in data:
to:
if 'ReferencedAction' in data and data['ReferencedAction'] is not None:
That’s it — just adding that extra check prevents the error.
I’ll publish the corrected script to the GitHub repository in the coming days.
Thanks again for sharing your files and helping me debug this!
P.S. I used Claude Opus 4.5 to debug this. Otherwise, all of this would be completely over my head.
Super @Kirk_Masden! Thanks for fixing this issue.
Now, it looks like I have a lot of cleanup to do. 
Thank you for your help Kirk!
I cannot wait to dig into the results.
@shimodabt and all,
In a private email to me, @shimodabt kindly provided feedback about false positives for orphan actions (actions that were actually in use) and pointed out some other issues that require fixing. I hope to be able to address these issues within a week or so.
Little by little, with feedback from people who use the scripts to reduce cruft and address other issues, I hope to make them more robust and effective.
Thanks to the data that @shimodabt kindly shared with me, I was able to do the following:
Bug Fixes:
Improvements:
I spent a lot of time debugging with my own apps but @shimodabt’s configurations were different from mine, so they helped me find problems I was unaware of. I hope that others of you who encounter problems will let me know so that I can make the suite of scripts more robust.
The GitHub repository has been updated: GitHub - KirkMasden/appsheet_parser_and_orphan_detector: Python tools for AppSheet developers: Convert HTML Application Documentation to CSV format, analyze component dependencies, and identify unused (orphan) elements for cleanup.
I also improved the documentation on my website:
Full documentation: AppSheet Parser Suite - Technical Documentation
Thanks again to @shimodabt! I suspect other issues will need to be addressed in the future. When that time comes, I will provide updates here.
I noticed recently that my app was running slowly. The performance monitor showed that one virtual field was the main issue. To improve speed, I replaced it with a sheet field using a formula. The challenge, of course, is remembering where that virtual field is used in the app so nothing breaks. To solve this, I used Kirk’s parser, which reported that the column field was referenced in four views. That insight let me quickly update those views with the new column field. Thanks @Kirk_Masden for the great tool — it saved me time and kept my app running smoothly!
Thanks so much, @shimodabt ! I’m very glad that my scripts have been of some help, even though they’re not as robust as they should be.
One of several things you helped me realize through our direct exchanges is that I should not have excluded settings tables (the system-created tables that allow individual users to control settings for their apps). In my own app building I had not used these and so I thought I could exclude them. Your app, however, showed me that they should be included and I am currently working on fixing that problem.
I’m rather slow. I have other work to do and the script adjustments often take me a while, even with the help of AI. But, I’m pretty confident that I can continue to make the scripts more robust (that is, get them to accommodate apps with configurations that I did not foresee). So, I hope that you and others will continue to post about problems you come across.
Thanks to issues that @shimodabt kindly alerted me to, I’ve made several improvements to the parser suite. These changes should improve results for creators who use User Settings in their apps. However, more testing is needed to make sure everything is working properly.
New Features:
USERSETTINGS("ColumnName") references, storing them as _Per User Settings[ColumnName] in the referenced_columns field. This enables proper dependency tracking for User Settings columns.USERSETTINGS(). Skips system columns (_EMAIL, _NAME, etc.), Show type columns, and unconfigured default slots (Option 1, Option 2, etc.).USERSETTINGS("X") references a column that doesn’t exist in _Per User Settings. This helped me find an actual bug in @shimodabt’s app, which I have been using for testing!Bug Fixes:
["MOCM","MASS"] were incorrectly captured as column references.CONTEXT("View") comparisons, matching AppSheet’s actual behavior. (AppSheet’s CONTEXT() and FIND() functions are the only case-sensitive functions.)The GitHub repository has been updated: GitHub - KirkMasden/appsheet_parser_and_orphan_detector
As always, please let me know if you encounter any issues!