You can do both:
Icon will serve for discoverability for first time users.
+
Visible and findable.
-
Small click target.
Double click is a productivity shortcut for advanced users.
+
Large click target (whole row), fast.
-
Invisible, needs explanation.
Don't forget a way to convert people from beginners to advanced users - for example, by a clue (row does react on hover) and by info (add a note to edit window opened by button 'Note: you can also edit rows by double clicking them.' or to button's title/tooltip).
EDIT: Also, what's the typical action? Will the user want to edit the whole record, or to fix a single attribute? If they often need to fix a single number, double clicking the table cell could do a fast inline edit (replace the table cell with input, save on <enter>
, clicking outside, or on a small inline OK button).
Observe what people do, make a shortcut to tighten their interaction loops. For example:
Before:
- find an edit button for a record
- click it
- wait for a window to open
- orient myself in an opened detail (which is probably laid out vertically instead of horizontal row I was opening)
- find a correct item
- focus it
- fix the value
- find save button
- click save
After:
- point at number I want to change
- double click
- fix the value
- hit enter or click save