Directory loader for bulk documents in LangChain - Performance & Optimization

Performance: Directory loader for bulk documents

MEDIUM IMPACT

This affects the initial page load speed and responsiveness by controlling how many documents are loaded and processed at once.

Loading many documents from a directory to process in bulk

LangChain

from langchain.document_loaders import DirectoryLoader
loader = DirectoryLoader('docs')
for doc in loader.lazy_load():
    process(doc)

Loads documents one by one lazily, allowing processing and rendering to start earlier.

📈 Performance Gainreduces blocking time, improves LCP by streaming documents

Loading many documents from a directory to process in bulk

LangChain

from langchain.document_loaders import DirectoryLoader
loader = DirectoryLoader('docs')
docs = loader.load()

Loads all documents at once, blocking the process until all files are read and parsed.

📉 Performance Costblocks rendering for hundreds of milliseconds or more depending on file count and size

Performance Comparison

Pattern	DOM Operations	Reflows	Paint Cost	Verdict
Load all documents at once	N/A (backend load)	Blocks main thread causing delayed reflows	High paint cost due to delayed content	[X] Bad
Lazy load documents one by one	N/A (backend load)	Minimal blocking, allows incremental reflows	Lower paint cost, faster content display	[OK] Good

Rendering Pipeline

Bulk loading all documents triggers a long blocking task before rendering can proceed. Lazy loading breaks this into smaller chunks, allowing the browser to interleave rendering and processing.

→Script Execution

→Layout

→Paint

⚠️ BottleneckScript Execution blocking main thread

Core Web Vital Affected

LCP

This affects the initial page load speed and responsiveness by controlling how many documents are loaded and processed at once.

Optimization Tips

1Avoid loading all documents at once to prevent blocking rendering.

2Use lazy loading or batch processing to improve initial load speed.

3Monitor main thread blocking time to optimize document loading.

Performance Quiz - 3 Questions

What is the main performance issue when loading all documents from a directory at once?

A It blocks the main thread causing delayed rendering

B It increases CSS selector complexity

C It causes layout shifts after rendering

D It reduces network requests

DevTools: Performance

How to check: Record a performance profile while loading documents; look for long scripting tasks blocking rendering.

What to look for: Long tasks in the main thread timeline indicate blocking; shorter tasks with gaps show better lazy loading.

Practice

1. What is the main purpose of the DirectoryLoader in Langchain?

easy

A. To convert documents into text format

B. To save documents to a directory

C. To load multiple documents from a folder at once

D. To create a new directory for documents

Solution

1. Step 1: Understand DirectoryLoader's role

   The DirectoryLoader is designed to read many files from a folder in one go.

2. Step 2: Compare options with this role

   Only To load multiple documents from a folder at once describes loading multiple documents from a folder, matching DirectoryLoader's purpose.

3. Final Answer:

   To load multiple documents from a folder at once -> Option C

4. Quick Check:

   DirectoryLoader loads many files = A [OK]

Hint: DirectoryLoader reads many files from a folder at once [OK]

Common Mistakes:

• Confusing loading with saving files
• Thinking it converts file formats
• Assuming it creates directories

2. Which of the following is the correct way to specify a glob pattern to load only text files using DirectoryLoader?

easy

A. glob='*.txt'

B. glob='text/*'

C. glob='**/*.doc'

D. glob='*.pdf'

Solution

1. Step 1: Understand glob pattern for text files

   The pattern '*.txt' matches all files ending with .txt in the directory.

2. Step 2: Evaluate each option

   glob='*.txt' correctly uses '*.txt' to select text files. Others select different or incorrect file types.

3. Final Answer:

   glob='*.txt' -> Option A

4. Quick Check:

   Use '*.txt' to pick text files [OK]

Hint: Use '*.txt' to select text files with glob [OK]

Common Mistakes:

• Using folder names instead of file patterns
• Using wrong file extensions
• Confusing recursive patterns with simple glob

3. Given the code snippet:

loader = DirectoryLoader('docs', glob='*.md')
docs = loader.load()

What will docs contain?

medium

A. A list of all markdown (.md) documents in the 'docs' folder

B. A single markdown document from the 'docs' folder

C. An error because glob pattern is incorrect

D. A list of all files in 'docs' regardless of extension

Solution

1. Step 1: Analyze DirectoryLoader usage

   The DirectoryLoader is set to load files from 'docs' folder with glob '*.md', meaning markdown files only.

2. Step 2: Understand load() output

   The load() method returns a list of documents matching the glob pattern, so docs will be a list of markdown files.

3. Final Answer:

   A list of all markdown (.md) documents in the 'docs' folder -> Option A

4. Quick Check:

   load() returns list of files matching glob [OK]

Hint: load() returns list of files matching glob pattern [OK]

Common Mistakes:

• Expecting a single document instead of a list
• Assuming glob pattern is ignored
• Thinking load() returns file paths instead of documents

4. What is wrong with this code snippet?

loader = DirectoryLoader('files', glob='*.txt')
docs = loader.load

medium

A. Incorrect glob pattern for text files

B. Missing parentheses after load method call

C. DirectoryLoader requires absolute path only

D. load method does not exist in DirectoryLoader

Solution

1. Step 1: Check method call syntax

   The load method must be called with parentheses: load()

2. Step 2: Verify other parts

   Glob pattern '*.txt' is correct, DirectoryLoader accepts relative paths, and load() method exists.

3. Final Answer:

   Missing parentheses after load method call -> Option B

4. Quick Check:

   Method calls need () to execute [OK]

Hint: Remember to add () when calling methods [OK]

Common Mistakes:

• Forgetting parentheses on method calls
• Assuming glob pattern is wrong
• Thinking DirectoryLoader needs absolute paths

5. You want to load all PDF and text files from a folder named 'resources' using DirectoryLoader. Which code snippet correctly achieves this?

hard

A. loader = DirectoryLoader('resources', glob='*.pdf;*.txt')

B. loader = DirectoryLoader('resources')
docs = loader.load() # loads all files

C. loader = DirectoryLoader('resources', glob='*.pdf|*.txt')
docs = loader.load()

D. loader = DirectoryLoader('resources', glob='*.pdf')
loader2 = DirectoryLoader('resources', glob='*.txt')
docs = loader.load() + loader2.load()

Solution

1. Step 1: Understand glob pattern limitations

   DirectoryLoader's glob does not support multiple patterns separated by ';' or '|'.

2. Step 2: Check each option

   loader = DirectoryLoader('resources', glob='*.pdf') loader2 = DirectoryLoader('resources', glob='*.txt') docs = loader.load() + loader2.load() loads PDFs and text files separately and combines results, which works correctly. loader = DirectoryLoader('resources') docs = loader.load() # loads all files loads all files but not filtered. Options A and C use invalid glob patterns.

3. Final Answer:

   loader = DirectoryLoader('resources', glob='*.pdf') loader2 = DirectoryLoader('resources', glob='*.txt') docs = loader.load() + loader2.load() -> Option D

4. Quick Check:

   Use separate loaders for multiple file types [OK]

Hint: Use separate loaders and combine for multiple file types [OK]

Common Mistakes:

• Trying to use multiple patterns in one glob string
• Assuming glob supports ';' or '|' separators
• Loading all files without filtering