•Filter a column from a searchable list of its values with counts, a number range drawn over that column's own distribution, or a date panel with presets and a custom range. A column that can't be filtered says so instead of offering a control that does nothing. Each filter shows as a chip you can edit or remove, and a filter added from a field's profile applies to the grid, the pivot and the chart alike.
•Open a CSV or TSV file from the explore page. While Rowmill prepares it, you see real row and byte counts rather than a spinner. Cancel a load with the Cancel button or the Escape key, and open a different file while one is already loaded. A file that isn't CSV or TSV is refused with a plain explanation.
•When a file won't load, Rowmill tells you why. A malformed row is named by its line number, and a file too big for your browser's memory says so, shown the same way as any other failure. If a replacement file fails to load, the Sample badge stays in place so you know what you're still looking at.
•The grid draws only the rows on screen, so you can scroll a file of millions of rows and reach every one of them. Sort, filter and group from a column header. A sort shows its progress and holds back the controls it would compete with until it finishes. A sort that doesn't fit in memory is worked through in parts on your device, though a very large one can still run out of memory. Dates and timestamps show as dates, not long numbers.
•Build a pivot by dragging fields into Rows, Columns and Values, or with the keyboard alone, starting from the field list. Move fields between zones, reorder them within a zone, and switch a value between Total, Average and the other options from its chip. Sort the result from its column headers, with totals worked out for you. Hover over or focus a field to see a profile of what's in it. The result recalculates and is announced on every change, and the result table shows a limited number of rows and columns so a very wide or very tall pivot stays responsive.
•Export the pivot result you're looking at as CSV or XLSX. A read-only drawer shows the SQL behind it, if you want to see it. When a pivot needs more memory than your browser has, Rowmill explains why and offers two ways to narrow it.
•Messages stack rather than replacing each other, pause while you hover over one, and stay on screen as you move between pages.
•Turn a pivot result into a chart, and hover to read exact values. Values on different scales are drawn as separate small charts rather than squeezed onto one axis, and negative values are drawn below zero.
•Filter panels keep long column names and long values inside their borders, and a field's card closes once you've added that field.
•Every column is profiled while your file loads, including an approximate median for number columns. A column's distribution is worked out when you open its field card, so your rows appear sooner, and a column with a very large number of distinct values is counted up to a limit.
•A live counter shows how many bytes Rowmill has sent from your device. The privacy line on the loading screen and the claim on the landing page both read from that counter, so neither can say something the counter doesn't show.
•The privacy panel is a send receipt: a list of everything this browser has sent us, newest first. Rowmill counts how it's used without any analytics company: the counts go to rowmill.app itself. They cover actions such as opening the SQL drawer or rebuilding a pivot, and whether a file came from your disk or the built-in sample. Nothing from inside your file is ever sent, not its name, its columns or a single value, and the site runs no script from anyone else.
•Save as recipe isn't built yet. It will let you drop next month's export and get the same filters, pivot and chart back. The button says so, and you can leave your email address to hear when it's ready.
•Send us feedback from inside the app. Before it goes, you see exactly what will be sent, and it travels through rowmill.app rather than opening your email program.
•The engine that reads your file runs inside your browser and is downloaded from rowmill.app, not from a third party. It starts the first time you ask it for something, shows a single full-screen progress view while it gets ready, and keeps running when you switch tabs or the browser sets the page aside. Cancelling stops the work rather than leaving it running. A second Rowmill tab tells you the first one is holding the file, and recovers on its own when that tab closes. When something fails, the message gives the reason in plain words, and the download size shown is what was actually transferred.
•A landing page, privacy and contact pages, and a 404 page in the same style. Each page has its own title and description for search results and link previews, and arrives as finished HTML, so it shows up before the scripts load. Moving between pages, including from the footer and the Open a file link, doesn't reload the page, and Ctrl-click or Cmd-click opens a link in a new tab.
•The landing page says what each figure measures: the width ladder names the operation it times, and the byte figures say what they cover. The recipe card describes the feature as it stands today.
•Try Rowmill without a file of your own. The sample is generated on your device when you ask for it, so there's nothing to download and nothing to upload.
•The waitlist form passes your address to our mailing list at Resend, and Rowmill keeps no copy of its own.
•The Rowmill mark appears across the app, the browser tab icons are drawn from it, and a Beta tag beside the name in the header says what stage this is at.
•Every control shows a visible ring when you reach it with Tab, and the edges of controls, zones and chips meet the 3:1 contrast minimum for non-text elements. In a window narrower than 900 pixels the layout rearranges to fit, and the SQL drawer opens from an icon button.
•Each release has a version number and a changelog, which is what you're reading.