Proposal: Standardizing "parse-" Commands in VisiData

[frosencrantz]

Proposal: Standardizing "parse-" Commands in VisiData

I'd like to propose establishing a standard protocol for commands that parse column values, similar to how "addcol-" and "setcol-" commands follow consistent behavior. The idea is to create a family of "parse-" commands that handle different types of structured data while maintaining consistent behavior.

Current Example

The existing addcol-split command provides a good model for this protocol:

• Creates a new column containing the parsed value as a Python object, a list. Other types might use a dict.
• Appends an identifying suffix to the new column name
• Preserves the original column

Proposed Extensions

These commands could have variants to:

• Optionally hide the original column
• Expand complex objects into multiple columns
• Filter which components of the parsed value to include

Potential Parse Commands

Here are some parsing targets that could follow this pattern:

1. Pattern-separated values (existing addcol-split)
2. File paths (leveraging pathlib properties)
3. VisiData keystroke sequences
4. URLs (using urllib.parse)
5. Dates and timestamps
6. JSON structures
7. Currency values (amount + symbol)
8. GitHub/GitLab/etc repository references
9. Semantic versions
10. GPS coordinates
11. Color specifications (RGB, HSL)
12. Street addresses
13. Person names

Alternative Approaches

We could also consider:

1. Extending the column type system building on how the existing currency and URL types work.
2. Integrating with expand- commands
3. Treating columns as objects with properties that can be extracted or combined

I'm interested in community feedback on:

• Additional parsing targets that would be valuable
• Preferred approach (parse commands vs column types)
• Specific behavior patterns these commands should follow
• Examples already in VisiData or in plugins.

This could be a great opportunity for new contributors to add parsing capabilities for different data types while following a consistent interface.

This post was solely aided by AI, and no mind-altering substances were harmed.