Name Splitter

The last space is the wrong place to cut

Split on the last space and John Smith comes out right. So does Mary Jones, and so do the first few hundred rows of almost any list, which is exactly why the problem goes unnoticed. Then you reach Ludwig van der Berg and the family name is Berg, with van der sitting in the middle name column as though it were a second forename.

And it is never only one row.

Dutch and Flemish lists are full of van, van der and van den. German ones have von and zu. Spanish has de, del and de la, and often two family names after it. Portuguese has dos and das. Arabic has bin, ibn and al. Irish has Mac and O'. Italian has di and della. French has de and le. Around forty of these words are carried here, and a family name that starts with one of them is kept whole.

And the page tells you when it did it. A row where words were joined comes back marked as likely rather than sure, with the words it joined named in plain English, because that call can go the other way and you are better placed than we are to know.

One column, two ways of writing a name

Export a list from one system and you get John Smith. Export from another and you get Smith, John. Put the two together, which is what happens the first time anybody merges a CRM with a spreadsheet, and you have a column holding both.

Nearly every tool decides once, for the whole column, usually by the majority. The minority then comes out backwards, and nothing says a word. Six months later somebody sends a mail merge that opens with Dear Smith.

Here each row is read by its own shape. A comma means the family name was written first, whatever the rest of the file looks like. The page counts how many of each it found and says so, because a column holding both is usually a sign that two lists were joined and there may be more than names to check.

Every row checked by putting it back together

This is the rule the whole page is built on. Take the parts it produced, join them back in order with the spacing they had, and you must get the original name back exactly, character for character. Not nearly. Exactly.

It sounds obvious and it is the check nobody runs, which is why dropped particles and lost nicknames go unnoticed. A name that loses de la still looks like a name. A row where a double space collapsed still looks fine. The only way to catch either is to compare what came out with what went in, on every row, every time.

So that is what happens, and any row that fails it is reported rather than passed off as fine. There is a second check underneath: the same name appearing twice in one file has to come out the same both times. If it ever does not, the page says so, because a parser that disagrees with itself is worse than one that is simply wrong.

Not everything in a name column is a person

Real lists have Acme Ltd in the name column. They have Northside Clinic and Hartley Consulting. Splitting those into a first and last name gives you a customer called Acme whose surname is Ltd, and a mail merge that says Dear Acme.

So around fifty words that mark a company are watched for, in several languages, and a row carrying one is marked unsure with the reason given rather than split as though it were a person. The row is still there and still whole. Nothing is dropped, because that decision is yours.

The same goes for two people in one cell. John & Jane Smith is one row and two people, and no set of columns holds that properly. It is marked, the row is kept intact, and you decide whether to split it into two rows before or after.

Finding the column, and the comma that is not a separator

With a header row, the column is found by its name, in twenty five languages: Name, Full Name, Contact, Nombre, Nom, Naam, Nome, Imię, 名前, الاسم. Without a header, it is found by shape. A column of emails is not a column of names. Neither is a column of numbers, web addresses, or sentences sixty characters long. Whichever it picks, it says so and you can change it.

And one that catches people out. A file holding a single column of names written Last, First, with no quotes around them, looks like a two column file to every reader ever written, because the comma is the separator. Half of every name ends up in a column nobody looks at. Here, a header with one cell and rows with two is taken as a sign that the comma belongs to the name, and the name is put back together before anything else happens.

What has actually been tested

Every figure here is a test run that has to pass before anything reaches this page. If one fails, nothing ships. A test reads this page, pulls out every figure in it, and compares each one with the code behind it.

• 60Names themselves, every shape, both orders, and the rule that nothing is ever lost names
• 41Every file that ships, scanned for generated copy, inline styles, off-origin loads and broken asset paths ship
• 34Every shape against every order, and the sizes pushed until the machine is the limit limits
• 37Finding the column, naming the new ones, and carrying your own columns through untouched columns
• 28Whole awkward files: no header, ragged rows, broken quotes, UTF-16, workbooks files
• 16The live page in a real browser, under the security policy it ships with e2e
• 17Every control used in a real browser, to prove none of them is decoration ui
• 13Layout at seven screen sizes, with the browser text enlarged, and the contrast of every line visual
• 13Every number printed on this page, checked against the code behind it figures
• 10Every icon and image is the real brand asset, and the link preview belongs to this tool brand
• 8The writing itself: sentence rhythm, repeated openers, and the templates that give generated prose away humanise

Properties you can hold it to

• It never loses a characterThe parts rejoin into the original name exactly, and every row is checked that way.
• It says how sure it isThree levels, and every row that is not sure carries its reason in plain words you can read.
• It agrees with itselfThe same name twice comes out the same both times, and a contradiction is reported as something to stop for.
• It reads each row by its own shapeBoth orders can sit in one column and neither gets reversed to suit the other.
• It leaves your file aloneYour columns come back exactly as they were and the new ones go on the end, never over the top of one of yours.
• It knows a company when it sees oneAnd a cell with two people in it, and says so instead of splitting them as though they were a person.
• It keeps the rows nobody asked aboutBlank cells stay in place rather than being dropped, so your rows still line up with everything else you have.
• It holds nothingNo account, no history, no storage. A list of people's names should not be passing through anybody else's server.

Questions people ask first