Enhanced conversions address_info normalization - inconsistent whitespace handling between docs and language samples

[frankii91]

Enhanced conversions address_info normalization - inconsistent whitespace handling between docs and language samples

Hi,

while implementing Enhanced Conversions with ConversionUploadService.UploadClickConversions, I ran into another documentation inconsistency related to normalization before hashing address_info fields.

This is similar to my previous normalization-related issue:

#1119

1. Whitespace normalization: docs vs language samples

On this Google Ads API documentation page:

https://developers.google.com/google-ads/api/docs/conversions/upload-offline#normalize-and-hash-user-provided-data

the "Normalize and hash user-provided data" section says, paraphrased:

• remove leading and trailing whitespace
• convert text to lowercase
• format phone numbers according to E.164

This appears to imply that internal spaces should remain.

However, the language samples on the same page behave differently:

• Java removes all whitespace with replaceAll("\\s+", "")
• PHP removes spaces with str_replace(' ', '', $normalized)
• Perl removes all whitespace with $value =~ s/\s+//g
• C# uses only Trim().ToLower()
• Python uses only strip().lower()
• Ruby uses only strip.downcase

Documentation page with code examples:

https://developers.google.com/google-ads/api/docs/conversions/upload-offline#code-example

2. Why this matters

For fields like:

• hashed_first_name
• hashed_last_name
• hashed_street_address

the hash changes depending on whether internal spaces are preserved or removed.

Example input:

ul. Jana Pawła II 10

Depending on the sample, the normalized value before SHA-256 hashing becomes:

ul. jana pawła ii 10

or:

ul.janapawłaii10

These produce different hashes.

3. Question

Which normalization is correct for enhanced conversions uploaded through Google Ads API, especially for UserIdentifier.address_info fields?

Should internal whitespace be preserved, as suggested by the main documentation and by the C# / Python / Ruby samples?

Or should all whitespace be removed, as done by the Java / PHP / Perl samples?

If one behavior is correct, could the documentation and all language samples be aligned?

Thanks.