undefined | Better HN

0 pointsjolmg4y ago0 comments

> though some seem to dislike it significantly

I can see why. The indentation of the whole statement is not determined by the first line, but by the 6th on the first and the 8th on the second on a `JOIN` clause. It's really arbitrary, and when you have that statement between other code, it's going to be weird how the start of the statement is much more indented than its preceding code. I really dislike it, too.

I prefer the use of indentation to signal what's inside another syntax structure. So, for example, I also dislike how you aligned `ON` and `AND` when the `AND` is inside the `ON` expression. It makes it seem like the two lines are on the same syntactic level.

Here's how I do it:

  SELECT a.foo
       , b.bar
       , g.zed
    FROM alpha a
    JOIN beta b  
      ON a.id = b.alpha_id
        AND a.another = b.thing
    LEFT JOIN gamma g 
      ON b.id = g.beta_id 
    WHERE a.val > 1
      AND b.col < 2
    ORDER BY a.foo

You might also notice that I removed the padding you used to align the `=` signs. I dislike big changes where the only thing that changed for a line is the whitespace padding. It obscures the real change. It might not seem like a big thing when you only have 2 lines in alignment, but it's a real bother when reading a diff that does that for more lines. You have to compare between the - and + lines to find what really changed instead of the diff telling you outright.

0 comments

dspillett4y ago

> and when you have that statement between other code

In many circumstances (chunks of SQL in SSIS steps and so forth) there is no code to be around. Within longer tracks of SQL I'll not let "LEFT OUTER JOIN" push the whole statement to the right, but will drop the table name to the next line and maintain the alignment from there on.

> removed the padding you used to align the `=` signs. I dislike big changes where the only thing that changed for a line is the whitespace padding. It obscures the real change

For modifying existing code, I will sometimes let the alignment slide to avoid unnecessary extra lines of change. Though a good visual diff tool will have the option to ignore white-space-only changes and I'm not concerned about patch sizes being a few lines bigger.

To fix the alignment there may be a subsequent check-in that tidies up non-functional elements of the code. I'm happy to have an extra item in history in order to maintain readability over a stream of unaligned text.

Something I've wanted for a long time in code editors is for tabs after the first non-whitespace character on a line to line up, that way such alignment could be managed without extra effort or extra lines in a diff. There would need to be some simple heuristics to control breaking the alignment (and these may need to vary between languages) and perhaps some way to control/override them (a directive in comments?) in edge cases.

j / k navigate · click thread line to collapse