Great question—let’s unpack what’s going on here, and fix your approach to be more reliable.

Why dir(cecmod) returns column names (and why that’s not its purpose) #

First, let’s clarify: pv.pvsystem.retrieve_sam('CECMod') returns a pandas DataFrame where each column represents a solar module model.

When you call dir() on a DataFrame instance (like cecmod), you’re asking Python to list all attributes and methods of that object. Pandas DataFrames have a handy (but easy to confuse) feature: they let you access columns as attributes, e.g., cecmod.Trina_Solar_TSM_300DEG5 (if that column exists). To make this work, pandas dynamically adds column names as "virtual" attributes to the DataFrame instance. That’s why those names show up in dir()—it’s a side effect of pandas’ attribute-based column access, not what dir() is designed for.

On the flip side, dir(pandas.DataFrame) (calling it on the class, not an instance) returns the class-level attributes and methods (like __init__, mean, groupby), which is what you’d normally expect from dir(). The difference between class and instance is key here.

Is using dir() a correct approach? #

No—while it might work in this case, it’s unreliable. dir() will include all sorts of other attributes and methods (like __len__, sum, isna) alongside column names. If any column name happens to match a built-in DataFrame method (unlikely for module names, but possible), you’ll get false positives. It’s not the intended way to access column names.

Better ways to get column names & filter for "Trina" #

Pandas provides dedicated tools for accessing column names—use these instead:

1. Use df.columns (most common) #

cecmod.columns returns a pandas Index object containing all column names. You can convert it to a list and filter with a comprehension:

# Filter columns containing "Trina"
matching_modules = [col for col in cecmod.columns if "Trina" in col]

2. Use pandas’ string filtering (more concise) #

Pandas has built-in string methods for Index objects, which makes filtering cleaner:

# Get all columns with "Trina" in the name, converted to a list
matching_modules = cecmod.columns[cecmod.columns.str.contains("Trina")].tolist()

3. Use df.keys() #

cecmod.keys() works exactly like cecmod.columns for DataFrames, so this is also valid:

matching_modules = [col for col in cecmod.keys() if "Trina" in col]

Fixed full code example #

Here’s how your code should look with the correct approach:

import pandas as pd
import pvlib as pv

# Retrieve the CEC module database
cecmod = pv.pvsystem.retrieve_sam('CECMod')

# Filter for modules with "Trina" in the name
matching_modules = cecmod.columns[cecmod.columns.str.contains("Trina")].tolist()

print(f"Found {len(matching_modules)} Trina modules:")
print(matching_modules)

Final takeaway #

• dir() is a general tool for inspecting object attributes, not for accessing DataFrame column names.
• Stick to df.columns or df.keys() for reliable column name access—these are purpose-built for this task.
• Pandas’ string methods make filtering column names (or any text data) much cleaner than manual list comprehensions.

内容的提问来源于stack exchange，提问作者GlenS