DATUI-KEYS(7) Miscellaneous DATUI-KEYS(7)

datui-keys - the keys of every datui screen

? or F1 on any screen shows that screen's keys, grouped by task: / narrows them to those whose text matches, and Enter closes the help and presses the key on the line. This page lists every screen's keys, with the longer descriptions.

? / F1
This screen's keys. In a text field ? types, and F1 opens the help
The home screen, on the dataset left, its filter kept and selected (typing replaces it); abandons a load
Quit from anywhere; Ctrl+C quits from a text field too, where Alt+W copies
Focus a form's field and act as Space: a checkbox flips, a choice steps, a picker opens, a button runs; a text field takes the cursor. A row of a list (a sort, a filter, a column) takes focus on the first click and acts on the second. A click on a picker's line chooses it, on a tab switches to it, and on a key in the footer presses it. A click outside a dialog does nothing. The mouse is the terminal's with display.mouse = false (or --mouse false); in most terminals Shift+drag selects text either way
Step a choice back, as ← does; else focus it

↑ / ↓ (j/k)
Move between the keys
Close the help and press the key on the line, where help was opened
/
Narrow to the keys whose text matches
Clear the filter, then close

Where a dataset opens.

Explore

↑ / ↓ (j/k)
Move the row cursor
← / → (h/l)
Move the column cursor, frozen columns included; the columns scroll only when it would leave the screen
A page of columns left or right, the cursor on the page's first column
{ / }
First column, last column
A page up or down (Ctrl+B / Ctrl+F too)
Half a page down or up
First or last row (G = End)
:
The command line: digits go to that row (the prefix says row:), anything else runs as SQL or q, as the prefix says; Ctrl+T switches. With a query in effect it opens on the query's text
Go to a column by name
/ (f)
Find text, a regex, or letters in order in the view; matches on screen light up as you type, and Enter takes the cursor, column cursor and all, to the first at or after its row. f opens it too
Next / previous match from the cursor's cell, wrapping round the view
On a row of a by query or a SQL GROUP BY, drill down to its rows (Esc comes back); elsewhere, inspect the row
Inspect the row: every field, each value whole and exact (Esc or Space closes)

Shape

[ / ]
Sort by the cursor's column, [ ascending and ] descending, replacing the sort in effect; the same key again on that column removes it. The s sidebar adds secondary sorts
Open the Sort & Filter sidebar (tabs: Columns, Filters), on the cursor's column
+ / -
Filter on the cursor's cell: + keeps the rows with its value, - drops them (a null cell: the nulls). Each adds a filter to the Sort & Filter sidebar, joined with "and"; the value is the cell's exactly as stored
Reverse sort order (sorted columns carry a direction mark in the header); with no sort, reverse the row order
Move the cursor's column one place left or right, the cursor with it; a frozen column moves among the frozen ones. R puts the order back
Pivot or melt
Draw a sample of the source into memory: the step under the query, filters and sort, which then run over it. The rows show as they arrive (Esc stops, keeping them); the footer says sample 100,000 of 36.8M. Analysis, charts and export read the sample. S again edits it; Method No sample, or R, takes it away
Reset table: clear the sample, query, filters, sort, column order, hidden columns and widths, frozen columns, pivot/melt, drill-down and the applied view

Analyze

Value counts of the cursor's column: each value's rows, percent and a bar, with a summary
Open Analysis. In a Data Quality evidence drill a is disabled; Esc returns to the observation
Chart the view
Open the Info panel (tabs: Schema, Resources, Partitions, Notes). H on its Schema tab reads a CSV's first row as data, or as column names

Output

Export the view to a file
Copy to the clipboard (cell, row, view or table); a cell is the cursor's
The saved views list
Apply the best-matching view (or open the list)

Display

#
Row numbers on or off: each row's place in the source, kept through a sort or a filter (a text file's line numbers). On for text and logs; display.row_numbers sets it
,
Number formatting (digit grouping) on or off
The type row under the headers on or off
< / >
The cursor's column 4 cells narrower or wider
= / w
Fit the cursor's column to the rows on screen; w puts it back to automatic width
A binary file read through a format spec: read it again with another spec. Clears the query, filters and sort
A file of several tables (a workbook's worksheets, a database's tables, a format spec's record types, a Hugging Face cache's splits): pick another to open in place of this one, as --table names it. Clears the query, filters and sort; a saved view for the table applies. On a file of one table, says so
Follow the file as it grows (CSV, TSV, PSV, NDJSON). Reads it again, so the query, filters and sort are cleared; while following, t pauses and resumes, and Esc stops

Go

Back to the home screen when the dataset was opened from it; otherwise quit
Quit
Leave a drill-down; stop a find, a sample being drawn (its rows so far stay) or a follow

Mouse

The cursor to the cell; on a header, its column
Enter on the row
Double-click a column's header to sort by it, as [ and ] do: ascending, then descending, then back to no sort, replacing the sort in effect
Double-click the gap right of a column's header to fit the column to the rows on screen, as = does; the column cursor moves to it, and nothing is sorted
↑ / ↓, three rows a notch; the same in help, the inspector and the sidebars
← / →, the column cursor (a sideways wheel too)
Press a key the footer shows; a click on the filters and sort presses s, on query presses :
Drag a column's header onto another column to move it there, as H / L do; a rule on the header marks where it lands
Drag the gap right of a column's header to set its width by hand, as < / > do, from 4 to 240 cells
Right-click a cell: the cursor goes there and a menu lists the keys that act on it (+, -, F, [, ], y, Space), each with its key. ↑ / ↓ and Enter, or a click, run a line as its key does; Esc or a click elsewhere closes it

datui with no path, or Ctrl+O from anywhere.

Explore

↑ / ↓
Move the selection (Ctrl+P / Ctrl+N too)
Previous or next section
A screenful, stopping at the first and last
The first or last row. The filter has no cursor to move: it is edited at its end
← / →
Fold or unfold a section; → on a directory or a file of tables goes inside it, and on a file a format spec reads as record types lists them, one row each
While the filter is empty: fold or unfold the section header under the cursor. With a filter typed, it types
Cycle the sort; the footer names the order in effect when it has room

Go

What the footer says on this row: "Open all" reads a whole directory as one table, "Inside" steps into it, "Open" loads a file, "Look" finds out first. A catalog bookmark, indented under its dataset, opens whole. On a section header, fold or unfold it; on the More row, show the rest; on the hidden-files row, show them
Delete a filter character; on an empty filter, up a level (from a bucket, back to its cloud source; from the top of a catalog's remote dataset, back here)
Back out one layer: the path prompt, the filter (onto the first dataset), the directory (back to the row it was entered from), then to the open table

Find

(type)
Narrow by name or column; fuzzy, so "sal" finds "sales". What you open often ranks first. Typing also searches below the directory you are inside and the listed bucket names; matches appear under "Found"
~
While the filter is empty: type a path or URL by hand. The list shows the directory being typed, narrowed by the name after the last /. The prompt is a plain editor: characters, Backspace, Ctrl+U clears, The first name that matches is picked. Tab completes the one name left or what the names share, or a name picked further down; ↑ / ↓ pick a name, and ↑ from the first takes the path as typed; Enter opens the picked name or the typed path: a file opens, a directory is gone inside; Esc closes. s3://, gs:// and az:// complete from buckets and prefixes already known. With a filter typed, ~ types into it
Clear the filter or path input
List again what is on screen

Manage

Show or hide files datui cannot read; inside a SQLite database, its internal tables
Show the local file under the cursor as bytes, in the hex view, whatever datui would read it as
Add the dataset or directory under the cursor to catalog.toml, listed under My datasets; on a row from catalog.toml, forget it. A heading stands for the directory it lists. Only catalog.toml is written; another catalog is hidden with home.hide
Open the Documentation view of a catalog row, of a place inside one, or of a file whose format spec documents it: description, publisher, license, links, record types, columns with units and value legends, bookmarks. A catalog's description, link and column notes stand over the spec's. Ctrl+E here is not readline's end of line: the filter is edited at its end
Forget the highlighted recent entry, or a whole place after confirming, or a row from catalog.toml, or hide a cloud source. On the Example datasets heading, hide them until datui cache clear, after confirming
Forget every recent entry, after confirming

Mouse

Select the row; double-click is Enter
Move the selection three rows, stopping at the ends

Ctrl+E on the home screen, on a catalog row or a file a format spec documents.

Read

↑ / ↓ (j/k)
Move the cursor a line
A page
The first or last line (Home / End too)
Open or close the value legend of the column under the cursor: a catalog's values, or a spec's enum
Open the link on the cursor's line in the system browser. Only http and https links open, and only after a question showing the whole URL (Enter opens, Esc does not). Off over SSH or without a display: y copies the link instead
Copy the link or value on the cursor's line, whole, however it is cut on screen
Back to the home screen

: at the table.

Run

(digits)
Digits alone go to that row, and the prefix says row: (:0 Enter is the top)
Go to the row, or run the query as the prefix says, sql: or q:. Reopened with a query in effect, the line holds its text, selected: typing replaces it, arrows edit it
Run, the same as Enter
Switch between SQL and q, keeping what is typed; the choice is remembered. [query] default_mode sets the first
Close

Edit

Complete a column name, or df in SQL; again for the next match. The line under the input lists the names that fit
SQL: start a new line
↑ / ↓
Earlier and later queries from the history (Ctrl+P / Ctrl+N too; SQL and q keep their own). In SQL over several lines, they move between lines first
Delete to the start or end of the line
Undo, redo

/ (or f) at the table.

Find

(text)
What to find: text (any case until a capital is typed), a regex with Ctrl+R, or letters in order with Ctrl+T. Matches in the rows on screen light up as you type, and the line says how many are on screen
Find: the cursor, column cursor and all, goes to the first match at or after its row, reading past the rows on hand when it must (the footer counts the rows read; Esc stops). On an empty field, clear the find
Keep only the rows with a match, as a filter: the footer and the Sort & Filter sidebar show it, and removing it there (or R) brings the rows back
Regex on or off
Letters in order on or off: smth finds Smith
Only the column cursor's column, or every column shown
↑ / ↓
Earlier patterns (Ctrl+P / Ctrl+N too)
Cancel

At the table

Next / previous match from the cursor's cell. Past the last match the find comes round to the first, and the footer says so. Each one typed while a find reads runs in turn; Esc stops them all
/
Find again, the last pattern ready to edit
Clear the find (or stop one still reading)

g at the table.

Go

(type)
Narrow to the names that contain it
↑ / ↓
Move
Go. The column cursor moves to it. A column already whole on screen stays where it is; another becomes the first after the frozen ones, or lands on the last page when it is there
Delete a character (Ctrl+W a word, Ctrl+U all)
Close without moving

Space at the table.

Fields

↑ / ↓ (j/k)
Move between fields
First and last field
A page of fields
← / → (h/l)
Previous and next row; the table's cursor moves with it
Into the value, to scroll and find in it. Nothing moves: the panes are split by what they hold, not by where the cursor is
On a group's row, its rows, as at the table. Else open a struct, a list, or text holding a JSON object or array; or read a field the table's rows do not hold (hidden and binary columns)
On a group's row, read a field the rows do not hold
/
Find a field by name, then by value: type to narrow, Enter or ↓ keeps the list narrowed, Esc clears it
Nulls: shown or hidden (null and empty fields); comparing, only the fields that differ
Order: the table's, A-Z, or nulls last
Compare: a column for the next row, or the pinned one
Pin this row to compare others with; again to unpin
Close. Esc backs out one level at a time: a find first, then Compare, then the inspector

Output

Copy the value as its view shows it
Copy the whole row as one JSON object
The value's next view, where it has more than one
Word wrap or hard wrap for long text
Open the value in another program

Value

↑ / ↓ (j/k)
Scroll a line
Scroll a page
Top and end, at once however long the value
/
Find in the value; n and N go to the next and last place
As in the fields
Back to the fields (Shift+Tab too); Esc clears a find first

Nested

Open the focused item
Into the item's value; not in an empty level
Up a level; at the row, Esc closes
Copy the focused item: text as itself, a JSON object or array indented

i at the table.

Panel

← / → (h/l)
Previous or next tab, from anywhere in the panel; Shift+Tab and Tab do the same. The panel is a viewer, not a form: its body always has the keys
↑ / ↓ (j/k)
Schema, Notes or Documentation tab: move the cursor. Model, Audio, MIDI, Metadata and format tabs: scroll the list
Model, Audio, MIDI, Metadata and format tabs: scroll the list a page
Model, Audio, MIDI, Metadata and format tabs: the top or the end of the list
Schema tab: change the column's type, with the names and formats a format spec's type takes. Notes tab: take the offer on the note, where it has one. Documentation tab: open or close the value legend of the column under the cursor. Excel or SQLite tab: open the worksheet or table under the cursor in place of this one, as T does
Documentation tab: open the link on the cursor's line in the system browser. Only http and https links open, and only after a question showing the whole URL. Off over SSH or without a display
Documentation tab: copy the link or value on the cursor's line, whole, however it is cut on screen
Schema tab, CSV, TSV, PSV: read the first row as data, under column_1, column_2, …; again to read it as column names. Reads the file again, so the query, filters and sort are cleared, and the panel closes
A dataset of more files than read.exact_count_files shows a row count estimated from a sample of its footers; c reads every footer for the exact count. The footer shows how far it has got, and Esc at the table stops it
Show the file's bytes in the hex view
Close the panel

F at the table.

Explore

↑ / ↓ (j/k)
Move
A page
First and last row
← / → (h/l)
The previous or next column; the table's column cursor moves with it
The rows holding the value, as a drill-down; Esc there comes back here
Sort by count or by value
Between a number column's histogram, binned from the counts, and the listing of its values. A number opens as its histogram
Count every row, when the counts are of a sample
While following a file, count again with the rows that arrived since; the bar says how many

Output

Copy the counts as TSV: every value, its count, percent and cumulative percent
Export the counts to a file
Back to the table; while every row is being counted, stop that and keep the sample

s at the table.

Sidebar

Next or previous field, wrapping; a field the form does not offer right now is skipped. The arrows move from the moment the dialog opens
← / →
On the tab bar: switch Sort & Filter / Columns. On a sort: flip its direction. On a filter: and / or. On a column: step its sort (none, ascending, descending). h/l too
On a sort: flip its direction. On a filter: edit it (column, operator, value). On "add sort": pick a column to sort by, last and ascending. On "add filter": a new filter, starting on the table's column cursor. On a column: step its sort
Apply everything staged and close, from any row (in the filter editor Enter takes the step; Ctrl+J applies)
Apply from anywhere, including mid-edit (the row in progress is saved). Ctrl+Enter does the same, on a terminal that tells it from Enter
Close an open picker or filter editor; otherwise cancel and close, discarding what is staged. Reopening shows what is applied

In effect

[ / ]
Move the focused sort earlier or later in the sort order, or the focused filter in the list
Remove the sort or filter
Remove every sort and filter

Columns

(type)
Narrow the column list, when the find field is focused. ↓ from find goes to the list, on the table's column cursor
A page of columns, or the first or last; ↓ stops at the last column
Cycle the column's sort: none → ascending → descending (← steps back). Each column carries its own direction
1-9
Put the column at that place in the sort order; 0 removes it (a digit past the end of the order says so)
Remove the column from the sort
[ / ]
Earlier or later in the sort order
+ / - (= / _)
Move the column's display position
Freeze this column and every column up to it at the left edge; on a column already frozen, pull the boundary back
Show or hide the column (it keeps its place, dimmed)
< / > (, / .)
Make the column 4 cells narrower or wider. A number column is never narrower than its numbers
Fit the column to the rows on screen, its name up to the automatic limit
Back to the automatic width
Clear the staged sort, order, locks, hidden columns and widths

Filter editor

(type)
Narrow the column or operator list. The operators are those the column's type takes
Pick the column (type to narrow, Enter chooses), the operator the same way, then type the value and Enter saves the row. Tab, → and Space also choose at the column and operator steps; Shift+Tab steps back; ↑ / ↓ move in the lists
End the edit, and only the edit

p at the table: the form on the left, a live preview of the result on the right (above and below on a narrow terminal).

Form

Next or previous field, wrapping; a field the form does not offer right now is skipped. The arrows move from the moment the builder opens
← / →
On the first row: switch Pivot and Melt. On the aggregation, strategy or type: the previous or next value. On a single column row: the previous or next column. In a text field: move the cursor. h/l too, outside text fields. The preview follows each change
On a column row: open its picker, scoped to that row; typing narrows it. On a choice: its next value, wrapping
Apply the reshape the preview shows, to the whole view, from any field
Close without applying; while a pivot is computed, stop it and keep the builder

Picker

(type)
Narrow the column list
↑ / ↓
Move; typing narrows
Choose; on a several-choice row, done
Choose; toggle a column on a several-choice row
Back out of the picker; the toggles made since it opened are undone (Enter keeps them)

c at the table: a chart chosen from the cursor column's type.

Shelves

1-7
Switch the chart type directly: Line, Scatter, Bar, Histogram, Box, KDE, Heatmap ([ / ] step). The shelves keep what the new type takes. On Rows, digits type a sample size instead
[ / ]
Previous or next chart type
Next or previous row of the panel, wrapping (j/k too). A shelf the type does not use is dimmed and skipped
On X, Y or Color: open its picker. On the line under Color: pick the values that get a series, by rows. On an option: toggle it or take its next value. The panel applies as it changes, so Enter acts as Space does, except on Rows: Space switches between Sample and Every row, and Enter reads what the row says
← / → (h/l)
Step the type, the time bucket (day, week, month, quarter, year), the aggregate (count, distinct, sum, mean, median, stdev, quantile, min, max, first, last) and a quantile's percentile, cumulative, bins, range or order; on Rows, switch between Sample and Every row (read on Enter); on a shelf that takes one column, the previous or next column; on the line under Color, turn Other (every value without a series) on or off; flip a toggle
+ / -
Bins or bandwidth
0-9
On Rows: type a sample size, like 50000, 50,000, 50k, 250k or 2m. Backspace edits, Enter reads it, Esc puts the row back as it was. A size of at least the table's rows is Every row. Nothing is read until Enter or focus leaves the row
Grid on or off at the labeled ticks: Line, Scatter, Histogram, Box and KDE. [analysis] chart_grid sets where it starts
Back to the table; on Rows with a change waiting, put the row back first

Plot

Line and Scatter: the plot takes the keys, and a crosshair reads out x and every series' value under the plot. ← / → (h/l) step to the next point or column, Home/End go to the ends; x, Tab or Esc hand the keys back to the panel. A click on the plot puts the crosshair there
Export the chart to PNG, SVG or PDF, with a title, notes and source. Needs the chart's shelves filled first
While following a file, draw again with the rows that arrived since; the bar says how many

Picker

↑ / ↓
Move; typing narrows
Choose; on a line or scatter chart's Y and on the Color values, Space toggles one in or out (up to 10, fewer on a terminal of fewer colors)
Choose and move to the next or previous row
Back out of the picker alone

Export dialog

Next or previous field: Path, Format, Style, Size, Width, Height, Legend, the marks (Opacity and Point size for a scatter, Line width for a line, Y from zero for a line), Title, Description, Notes, Source, Byline, Recipe
← / →
Step the format (PNG, SVG, PDF), the style (Light, Dark, Transparent), the size (Slide 16:9, Document, Square, Single column, Double column, Custom), the legend (line ends, a corner, off), a scatter's point opacity (Auto, 100%, 50%, 20%) and size (Small, Medium, Large), a line's width (Thin, Normal, Bold), Y from zero, or the recipe (Include, Omit). Typing a width or height makes the size Custom
Earlier or later paths in the path field
Export, from anywhere in the dialog. A path ending .png, .svg or .pdf takes that format; any other gets the format's extension after it. An existing file asks Overwrite / No, starting on No; ←/→ (h/l) or Tab pick, Enter confirms, and declining returns to the filled dialog
Back to the chart

a at the table.

Explore

Between the results and the tools
↑ / ↓ (j/k)
Rows, or the sidebar's tools
← / → (h/l)
Scroll the statistics; the header counts those out of view
First or last row
A page
Open the sidebar's tool and move into its pane. The first run on a dataset starts in the Sample form, where Enter runs it; later tools reuse that sample. Results last until the view changes: closing and reopening shows them again
Cancel a run in progress; otherwise from the results back to the tools, and from the tools close the analysis view

Sample

Choose the sample every tool reads: which rows, how they are picked, how many (typed, like 50000, 50k or 2m), the seed. Enter applies the form
View the sample's rows as a table; Esc comes back
Draw another sample (when the result is a sample)
Read every row instead, after confirming
While following a file, read again with the rows that arrived since the results were read

Distribution in the Analysis sidebar.

Explore

↑ / ↓ (j/k)
Rows, or the sidebar's tools
← / → (h/l)
Scroll the statistics; the header counts those out of view
First or last row
A page
Between the results and the tools
Open detail view for selected column (shows Q-Q plot and histogram); with the sidebar focused, open a tool
Cancel a run in progress; otherwise from the results back to the tools, and from the tools close the analysis view

Sample

Choose the sample every tool reads: which rows, how they are picked, how many (typed, like 50000, 50k or 2m), the seed. Enter applies the form
View the sample's rows as a table; Esc comes back
Draw another sample (when the result is a sample)
Read every row instead, after confirming
While following a file, read again with the rows that arrived since the results were read

Enter on a column in Distribution.

Detail

↑ / ↓ (j/k)
Compare the values with another family
The first or last family
Histogram scale: linear or log
Back to the distribution table

Correlation Matrix in the Analysis sidebar.

Explore

Between the matrix and the tools
↑ / ↓ (j/k)
Matrix rows, or the sidebar's tools
← / → (h/l)
Matrix columns
Jump to the first pair (the first row's second column) or the last (the last row's next-to-last column); the matrix opens on the first
A page
Open pair detail view (on a cell) or open a tool (sidebar); does nothing on a diagonal cell
Method: Pearson r or Spearman ρ, named in the title (both come from the one run, so it reads nothing)
Cancel a run in progress; otherwise from the matrix back to the tools, and from the tools close the analysis view

Sample

Choose the sample every tool reads: which rows, how they are picked, how many (typed, like 50000, 50k or 2m), the seed. Enter applies the form
View the sample's rows as a table; Esc comes back
Draw another sample (when the result is a sample)
Read every row instead, after confirming
While following a file, read again with the rows that arrived since the results were read

Enter on a pair in the correlation matrix.

Detail

Pearson or Spearman
Return to the correlation matrix. Resampling (r) works from the matrix, not from inside this detail

Data Quality in the Analysis sidebar.

Setup

↑ / ↓, Tab
Move between rows
← / →
Change a short list's choice
Open the row: the Sample form, a list (type to narrow), Time roles, Intervals, Column intent or Expected
The Sample form; its Enter applies to Setup and returns there
The access plan: what a run reads
Release the rows runs kept and a full scan's local copy, named on the Read rule; the next run that would reuse them reads again
Run, from any row. A full read asks first; the report on screen or in the session cache is shown, not read again
Discard every staged edit

Setup lists

↑ / ↓
Time roles: choose the role. Intervals: choose the pair. Column intent: choose the column. Expected: Windows, From, Before (Tab too)
← / →
Time roles: choose the role's column. Expected: choose the windows. Intervals: measure the pair or not
Intervals: measure the pair or not. Column intent: declare the column's intent in a form
Done
Put the list back as it was

Report

← / → (h/l)
Previous or next page
1 - 5
Overview, Columns, Segments, Trends, Intervals
↑ / ↓ (j/k)
Move, or scroll a tall finding
Page; Home/End jump to either end
Open a finding, then its rows. On an empty page, open the setting it needs
Overview: only one column's or one type's findings
Overview: ranked, by rows, by rate
Setup
Setup, with the Sample form open
View the sample's rows; Esc returns
The access plan: what a run reads
On a sample, run again with a new seed
Export the report to JSON or Markdown; nothing is read
Between the result and the tools
Back one level: all findings again, then the tools, then close

Segments

A segment's columns beside the one it is compared with, largest change first
Largest change first, or back in order
Compare with the highlighted segment

Trends

A line's bars: span, segments, rows sampled of counted, rate, 95% interval, and the bar before. ↑ / ↓ there: the next bar
Next measure
Stage a coarser window in Setup, for segments the sample reached thinly or not at all
The expected windows with no rows: empty by exact count, not sampled, or out of scope
Back to Trends

Intervals

An interval's detail: its ends, the rows with both, missing and unread ends, negative and zero durations, percentiles and breaches. Enter there shows the rows behind the count under the cursor: a sample's from the rows the run kept
Back to the list

e at the table.

Form

Next or previous field, wrapping; the fields follow the format. The dialog opens on Path, and the arrows move from there
← / →
On Format or Compression: the previous or next value, along the Format row (h/l too). On Header or Source file: toggle. In Path and Delimiter: move the cursor
Toggle a checkbox (Header, Source file); on Format or Compression, the next value, wrapping
In the path field: the paths exported to before, earlier or later (↑ and ↓ move between fields)
Export, from anywhere in the form. On a blank path the form says "Enter a file path." instead of exporting
Close without exporting

File exists

← / → (h/l)
Pick Overwrite or No (Tab too)
Confirm the one picked
Decline: back to the filled form

y at the table.

Form

Next or previous row
← / →
The previous or next scope, column or format (h/l too); on Header: toggle
On Scope or Format: the next value, wrapping. On Column: open its picker. On Header: toggle
Copy, from anywhere in the form. On the Cell scope with no column picked, Enter re-accents the spec line instead of copying
Close a picker, then the dialog, without copying

Picker

(type)
Narrow the list
↑ / ↓
Move the cursor (j/k narrow the picker; only ↑/↓ move there)
Choose
Choose and move on

v at the table.

List

↑ / ↓ (j/k)
Move in the list
Apply the selected view
Save the current state as a view (an untouched table has nothing to save, and s says so)
Edit the selected view
Delete the selected view, after confirming
Show how the selected view's score was computed (Esc closes the score view)
Close

Save and edit

Next or previous row. In the description ↑ / ↓ move between its lines, and leave it from the first or last
Save, from any row. In the description Enter types: Ctrl+J saves from there
Save from anywhere, the description included; so does Ctrl+Enter, on a terminal that tells it from Enter
Five lines in the description
Expand or collapse Matching; toggle schema match
Back to the list, discarding edits

Delete

← / → (h/l)
Pick Delete or No (Tab too); it starts on No
Confirm the one picked
Keep the view

b at a table read through a format spec.

Pick

(type)
Narrow to the names that contain it
↑ / ↓
Move
Read the file again with the spec chosen. The query, filters and sort are cleared
Delete a character (Ctrl+W a word, Ctrl+U all)
Close and keep the format

Enter on the Info panel's Schema tab, or Change type in the cell menu.

Pick

(type)
Narrow the list to the names that contain it. In the format list, a strftime format typed that no line holds is the format
↑ / ↓
Move
The type: the column reads as it at once, a value that does not fit null. A date, time or datetime asks its format next, each line showing what it makes of the column's first value. as read takes the type away
Delete a character (Ctrl+W a word, Ctrl+U all)
Back to the types, or close

Combine into datetime in the cell menu, on a text, date or time column.

Fields

Next or previous field
Pick a column, or step the kind
← / →
Step the kind: datetime, date or time
Make the column, before the first column it is made from, as a format spec's derived column is: a date and a time, and a UTC offset, make a datetime in UTC
Cancel

T at a table of a file of several.

Pick

(type)
Narrow to the names that contain it
↑ / ↓
Move
Open the table chosen in place of this one. The query, filters and sort are cleared
Delete a character (Ctrl+W a word, Ctrl+U all)
Close and keep the table

S at the table.

Form

Next or previous row
← / →
Step Rows from, Method and Per value of; Method's No sample takes the view's sample away
(type)
Type into the focused row: the size (50000, 50k, 2m), the seed, a row range, partition values or file numbers
Draw the sample: the table shows its rows as they arrive, and the query, filters and sort run over them. When the estimate is more than the memory available now (or analysis.sample_memory_limit), the form says so; Enter again draws anyway
Close; the view's sample stays as it was

datui --hex FILE, Ctrl+X on the home screen, or x in the Info panel.

Explore

← / → (h/l)
A byte
↑ / ↓ (j/k)
A row
The next group of four, or back one
0 / $
The start or end of the row
The start or end of the file (Home and End too)
A page (Ctrl+B/F too; Ctrl+U/D half a page)
:
Go to an offset: 4096 or 0x1000; +16 or -16 from the cursor; e-8 for the eighth byte from the end

Find

Find bytes. Text is found as its UTF-8 bytes (Ctrl+U in the prompt: as UTF-16 little-endian). 0x1acffc1d, or two or more hex pairs (de ad be ef), is a byte pattern, where ?? matches any byte. Text in double quotes is text, even when it looks like hex. A match may span rows
The next or previous match, round the end of the file
When the matches after the one found are all the same distance apart, make that the bytes per row
Stop a find that is reading

Display

Bytes per row, so that records line up; empty goes back to as many as fit. --hex-width N sets it on the command line
#
Offsets in decimal or hex
Show or hide the byte inspector. Beside the bytes when there is room for it and 16 bytes a row, over them when there is not
Mark from the cursor; move to mark a range, and the status line counts it. v again, or Esc, unmarks

Go

Read the file with a format spec instead
Back to the table, when opened from the Info panel; back home, when opened from there
Home, when opened from there; else quit

The keys of every screen, as a manual page.

datui man keys

datui(1)

The datui documentation: <https://derekwisong.github.io/datui/>

Report bugs at <https://github.com/derekwisong/datui/issues>.

Derek Wisong and the datui contributors.

Copyright © 2026 Derek Wisong

datui is free software under the MIT License.

2026-10-06 datui 0.4.1