git add

git add is the most frequent Git command after git status. It handles the transition from "edits on disk" to "edits ready to commit".

What happens under the hood

For each file:

1. Reads the content from the working tree.
2. Creates a blob object in .git/objects/.
3. Updates the record in .git/index: path, permissions, blob SHA.

After this, git status shows the file under "Changes to be committed".

Basic forms

git add file.txt          # single file

git add src/              # directory, recursively

git add .                 # everything from the current directory

git add -A                # all changes across the entire repository

git add '*.py'            # by glob pattern

The difference between . and -A: the dot adds only from the current working directory; -A looks at the whole repository regardless of where you are.

Interactive mode

git add -p file.txt

Splits the edit into hunks (chunks with ±3 lines of context) and asks about each one: y to add, n to skip, s to split into smaller pieces, e to edit manually. This is the primary tool for splitting one file that has two unrelated changes into two separate commits.

Pitfalls

• git add file.txt writes the current contents to the index. If you edit the file again after add, the old version stays in the index. git status will show the file under both "to be committed" and "not staged". Run git add again to catch up.
• New files are not tracked by default. They appear as untracked until you add them explicitly. There is no automatic registration. This is intentional: accidental files (logs, build artifacts) should not end up in the index.