enh: add expr.dt.to_string for SparkLikeExpr

[lucas-nelson-uiuc]

What type of PR is this? (check all applicable)

• 💾 Refactor
• ✨ Feature
• 🐛 Bug Fix
• 🔧 Optimization
• 📝 Documentation
• ✅ Test
• 🐳 Other

Related issues

• Related issue #<issue number>
• Closes #<issue number>

Checklist

• Code follows style guide (ruff)
• Tests added
• Documented the changes

If you have comments or can explain your changes, please do so below

Implementation for expr_dt.to_string - disclaimer: probably glossing over some dt/tz details, feel free to call me out

Makes use of the convention mapping from strptime_to_pyspark_format - can move this to _spark_like.utils for expr_dt and expr_str to import from:

narwhals/narwhals/_spark_like/expr_str.py

Lines 119 to 122 in dd41833

	def strptime_to_pyspark_format(format: str) -> str:
	"""Converts a Python strptime datetime format string to a PySpark datetime format string."""
	# Mapping from Python strptime format to PySpark format

couple notes:

• some pyspark formats aren't zero-padded - e.g. the format "%Y-%m-%d" should evaluate to "yyyy-MM-dd" in pyspark but this returns "y-M-d" which causes some minor differences (instead of 2020-01-09 we get 2020-1-9). wondering if we should update the pyspark mappings to their zero-padded equivalents?

narwhals/narwhals/_spark_like/expr_str.py

Lines 125 to 141 in dd41833

	format_mapping = {
	"%Y": "y", # Year with century
	"%y": "y", # Year without century
	"%m": "M", # Month
	"%d": "d", # Day of the month
	"%H": "H", # Hour (24-hour clock) 0-23
	"%I": "h", # Hour (12-hour clock) 1-12
	"%M": "m", # Minute
	"%S": "s", # Second
	"%f": "S", # Microseconds -> Milliseconds
	"%p": "a", # AM/PM
	"%a": "E", # Abbreviated weekday name
	"%A": "E", # Full weekday name
	"%j": "D", # Day of the year
	"%z": "Z", # Timezone offset
	"%s": "X", # Unix timestamp
	}

• not all tests pass for a couple reasons:
  • pattern contains week-based patterns (W): IllegalArgumentException: All week-based patterns are unsupported since Spark 3.0, detected: W, Please use the SQL function EXTRACT instead
  • timestamp handling: two things going on here:
    • similar to above, %f shortchanges the milliseconds by only returning one decimal place. could default to higher number of places
    • strptime_to_pyspark_format removes all instances of "T" for timezones

let me know if there's anything else i'm missing that I can implement

[FBruzzesi]

Hey @lucas-nelson-uiuc , thanks for the contribution.

There is an open PR (#1842) to add the same feature - we couldn't finish all the details in there, but would love some help and/or feedback to push it forward

[lucas-nelson-uiuc]

ah totally missed that - thanks @FBruzzesi I'll check it out

[FBruzzesi]

@lucas-nelson-uiuc thanks again for the effort here! I am going to close this one as I am trying to finalize #1842
Feel free to comment/review there if needed 🙌🏼