The spreadsheet: table formulas (#+TBLFM)

Almost every example comes twice:

1. A solved table. Its results were computed by org.nvim itself and pasted here, so recalculating it must change nothing. If a value changes when you press <C-c><C-c> on its #+TBLFM: line, you found a bug. (Tables that need several passes are marked with a comment # Solved with 16<C-c>*; one table in "Ranges" changes on purpose and says so.)
2. A practice copy right below it, announced by a comment line starting with # Practice copy. Its result fields are empty on purpose. You compute them and compare with the solved table and the Expect: text.

Comment lines (# ... at column 0) are notes for you; they are never part of a table or exported.

Only the first #+TBLFM: line under a table is its formula line. You can keep alternatives on more lines below it: <C-c><C-c> on any of them applies that line once, without making it the active one.

Try: in the practice copy press <C-c><C-c> on the second #+TBLFM: line.

Expect: squares: 1, 4, 9, 16.

Try: now <C-c><C-c> on the third line, then <prefix>Tf inside the table.

Expect: first 100%, 200%, 300%, 400%; then <prefix>Tf uses the first line again and gives 2, 4, 6, 8.

<C-c>* without a count recalculates only the column formulas of the row under the cursor (field formulas always run).

Try: put the cursor on the row | 4 | 5 | and press <C-c>*.

Expect: only that row gets a result, 20; the other two stay empty. Then press 4<C-c>*: 6 and 42 appear as well.

Formulas run top to bottom, column formulas first. In the table below the Share column divides each row by the total in the last row, but that total is a field formula, so it is computed after the shares. The first pass therefore divides by an empty field; a second pass uses the fresh total.

Try: in the practice copy press <prefix>Tf once.

Expect: nonsense shares 1200.0%, 800.0%, 2000.0% and 0.0% in the Total row: the total was still empty (an empty field counts as 0, and a division by 0 stays unevaluated, of which %.1f keeps only the leading number). The Hours total 40 is there now.

Try: press <prefix>Tf a second time.

Expect: 30.0%, 20.0%, 50.0% and 100.0% (the column formula also runs on the Total row). A third <prefix>Tf changes nothing.

Try: undo back to the empty copy (u a few times) and press 16<C-c>*.

Expect: the same final values in one go, and the message "Convergence after 3 iterations" (two passes that change something and a third that confirms nothing changes).

Note the order in 100*$2/@>$2. Written as $2/@>$2*100 it would be wrong: in Calc, / binds looser than *, so that means $2/(@>$2*100) (see "Arithmetic" below).

In the table below every number tells you where it lives: 23 is in row 2, column 3; 45 in row 4, column 5. The first column holds labels (and the header row is row 1).

How each result comes about:

• @2$5=@4$3 copies the field of row 4, column 3: 43.
• @3$5=$2+$4 uses columns 2 and 4 of its own row (row 3): 32+34 = 66.
• @4$5=@2*2 uses row 2 of its own column ($5), which the first formula just set to 43: 43*2 = 86.
• @5$5=@3$2+@4$4 = 32+44 = 76.

Try: press <C-c><C-c> on the practice copy's #+TBLFM: line.

Expect: out = 43, 66, 86, 76.

Try: change 43 (row 4, c3) in the practice copy to 100, recalculate with <prefix>Tf.

Expect: out = 100, 66, 200, 76: the first formula copies the new value, and @4$5 doubles it.

Try: press <C-c>} inside the table.

Expect: virtual text shows the row numbers (@1, @2 …) at the left of each row and the column numbers ($1, $2 …) above the table. <C-c>} again removes them. Horizontal lines get no number.

A signed number counts from the field being computed. They are most useful in column formulas, where the same formula runs on every row.

Here is a change column that compares each row with the row above it.

The left side @3$3..@>$3 is a range: the formula fills rows 3 to the last row of column 3. Row 2 (Sep) has no row above it with a number, so it is left out.

Try: recalculate the practice copy (<C-c><C-c> on #+TBLFM:).

Expect: Change = (empty), 150, -40, 290.

Try: change Nov's visitors to 1400 and press <prefix>Tf.

Expect: Change = 150, 50, 200.

These keep working when rows or columns are added, which absolute numbers do not. This table has no header, so row 1 is the first line.

• @<$< first row, first column: 11.
• @>$< last row, first column: 41.
• @<<$>> second row, second-to-last column ($3; $4 is the last): 23.
• @>>$2 second-to-last row, column 2: 32.

Expect: after <C-c><C-c> on the practice #+TBLFM:, column 4 reads 11, 41, 23, 32.

Try: on the row | 41 | 42 | ... of the practice copy press <M-CR> to add an empty row below it (the cursor moves there), press i and type 51 <Tab> 52 <Tab> 53, then <Esc>, and recalculate with <prefix>Tf.

Expect: @>$< now finds the new last row: 51; @>>$2 finds 42. The other two stay 11 and 23.

@>$> is the bottom-right corner, the classic place for a grand total.

$>…= is a column formula for the last column. It runs on the Total row too, but the field formula @>$> overrides it there.

Expect: Year = 95, 65; Total row = 75, 85, 160.

Each row compares its weight with the first row of its own section.

Expect: 0.0, -0.8, -1.1 in the first section and 0.0, -0.6 in the second: week 4 is the start of its own section.

Try: in the practice copy delete the middle horizontal line (dd on it) and recalculate.

Expect: now there is one section: 0.0, -0.8, -1.1, -1.9, -2.5.

@-1$-1..@+1$+1 is the 3x3 block around the field being computed (the field itself is empty, so it is dropped).

Expect: the centre field is 40 (1+2+3+4+6+7+8+9).

Try: press <C-c><C-c> on that #+TBLFM: line again.

Expect: 80! The field is no longer empty, so the second time it is part of its own block: 40 + 40. A formula that reads its own field changes on every recalculation. Such a table never becomes stable: 16<C-c>* gives up with "No convergence after 10 iterations". Press u to get 40 back.

One range formula fills a whole rectangle. @0$1 is column 1 of the current row, @1$0 row 1 (the header) of the current column.

Expect: the usual multiplication table; the bottom-right field is 25.

Try: change the header 5 to 10 and the row label 5 to 10, recalculate.

Expect: the last column becomes 10, 20, 30, 40, 100, and the last row 10, 20, 30, 40, 100.

The same table without any labels, only from @# and $#:

Expect: header 1 2 3 4, first column 1 2 3, and row r column c holds (r-1)*c, so the last row is 3 6 9 12. The three formulas fill the header, the first column and the rest; @1$1..@1$> is a range inside the header row, which column formulas never touch.

Expect: 10, 999, 30: the field formula for row 3 (item B) beats the column formula. The header x10 is never touched.

Try: remove ::@3$3=999 from the practice copy's formula line and recalculate.

Expect: 10, 20, 30.

Without a horizontal line, the column formula runs on every row.

Expect: 1, 4, 9.

Try: in the practice copy put the cursor on the first row and press <prefix>T- to insert a horizontal line below it. Recalculate.

Expect: the first row is now a header: its 1 stays, only rows 2 and 3 get 4 and 9.

A range on the left fills each field of the block; relative references inside it ($1, @-1) are relative to each field in turn.

Expect: b = 100, 200, 3000, 4000; c = 2, 4, 6, 8.

Formulas are applied in a fixed order: all column formulas (row by row), then field and range formulas sorted by their left side (row first, then column). A formula that reads a field computed by a later formula sees the old value. Here c adds a and b, but the range formula for c (@2$3..) sorts before the one filling rows 4-5 of b (@4$2..):

Try: press <C-c><C-c> on its #+TBLFM: line once.

Expect: b = 100, 200, 3000, 4000 but c = 101, 202, 3, 4: when c was computed for rows 4 and 5, b was still empty there.

Try: press <C-c><C-c> on it again (or 16<C-c>* from the start).

Expect: c = 101, 202, 3003, 4004, and further passes change nothing.

Expect: after <C-c><C-c> on the practice #+TBLFM:, a third column appears with 6 and 20; its header is empty (type one in if you like).

A field formula outside the table is an error instead (see the option table_formula_create_columns): try #+TBLFM: @2$4=1 on the copy and you get "Table formula error: Missing columns in the table. Aborting".

In Insert mode, type a formula into a field and leave the field with <Tab>, <CR> or <C-c><C-c>:

• =$1*2 becomes the column formula of that column ($2=$1*2).
• :=vsum(@I..@II) becomes a field formula for that one field.

The formula moves to the #+TBLFM: line (created if needed) and the field is computed at once. Other fields of the column are updated on the next recalculation (<prefix>Tf).

Try: put the cursor in the first empty field of "double" (row 2), press i, type =$1*2 and press <Tab>, then <Esc>.

Expect: a line #+TBLFM: $2=$1*2 appears under the table and that one field shows 2; the other rows are still empty.

Try: now go to the bottom field of the "double" column (the row below the second line), press i, type :=vsum(@I..@II), press <Esc> and then <C-c><C-c>. (In the last field of a table <Tab> would also add a new row, so <Esc> and <C-c><C-c> is the cleaner way here.)

Expect: the formula line becomes #+TBLFM: $2=$1*2::@5$2=vsum(@I..@II) and the field shows 2 (the only value computed so far).

Try: press <prefix>Tf.

Expect: 2, 4, 6 and 12 in the bottom field: the field formula @5$2 wins over the column formula there.

In Normal mode on a field:

• <C-c>= asks for the column formula ("Column formula $2=") and computes the field. The current formula is offered as the default; an empty answer removes the formula.
• 4<C-c>= (Emacs C-u C-c =) asks for a field formula for this field ("Field formula @3$2=").
• 16<C-c>= puts the formula that is active in this field into the field as =... (or :=...) so you can edit it in place and press <Tab>.

Try: on the first empty Fahrenheit field press <C-c>=, type $1*9/5+32 and <CR>. Then <prefix>Tf.

Expect: #+TBLFM: $2=$1*9/5+32 appears; the values are 32, 68, 98.6, 212.

Careful with Calc's precedence: / binds looser than *. $1*9/5 is ($1*9)/5, fine; but $1/5*9 would be $1/(5*9), which is wrong. Write $1*9/5 or use parentheses: ($1/5)*9.

<prefix>' in a table, or on its #+TBLFM: line, opens every formula of the table in a split below, one per line, sorted into sections. For the first table of this file ("A first spreadsheet") it shows:

# Column Formulas
$4 = $2*$3;%.2f

# Field and Range Formulas
@>$4 = vsum(@I..@II);%.2f

Edit freely, then:

While the cursor moves over a formula, the fields it refers to are highlighted in the table window, and the target field is highlighted too. A line starting with a blank continues the formula above, so long Lisp formulas can be split over several lines.

Try: put the cursor in this practice copy and press <prefix>'. Move the cursor over $2 and $3 of the first formula.

Expect: a split titled like the example above; the Qty and Price fields of one row light up in the table.

Try: in the editor change %.2f of the column formula to %.1f, and press 4<C-c><C-c>.

Expect: the editor closes, the #+TBLFM: line now reads $4=$2*$3;%.1f::@>$4=vsum(@I..@II);%.2f, and Total shows 1.5, 1.5 with Sum 3.00.

<C-c>{ toggles the debugger. While it is on, every formula evaluation opens a *Substitution History* window showing the steps and asks "Debugging Formula. Continue to next?". Answer y to go on, n to stop. For $2=$1*10;%.1f on a field 2 it shows:

Substitution history of formula
Orig:   $1*10
$xyz->  $1*10
@r$c->  (2)*10
$1->    (2)*10
Result: 20
Format: %.1f
Final:  20.0

• Orig is the formula as written, $xyz-> after names were replaced, @r$c-> after references were replaced by field values in parentheses.
• Result is the Calc result, Format the printf format, Final the text written into the field.

Try: put the cursor in the table, press <C-c>{ ("Formula debugging has been turned on"), then <C-c><C-c> on the #+TBLFM: line. Answer n to the question. Press <C-c>{ again to turn the debugger off.

Expect: the window shows exactly the history above. After n the recalculation stops: the field that was just computed keeps its 20.0, but the row with 3 stays empty. Recalculate normally to fill it (30.0).

Ranges may hold text; Calc keeps words as symbols, so vsum of a, 3, b is a + b + 3. That is rarely what you want, so keep text out of the ranges you sum.

<C-c>+ is not a formula: it sums the numbers of the current column (or of a Visual block) and puts the sum into the unnamed register, so p pastes it.

Try: put the cursor on a score of the practice copy, press <C-c>+, then :echo @".

Expect: the message "Sum of 7 items: 574" (the header and the empty field are skipped) and 574 in the register. Do it before you recalculate the practice copy: afterwards the result rows are numbers too and are added in.

Calc's trigonometric functions work in degrees unless you add the R flag (radians). D forces degrees.

Expect: sin 30 = 0.5000, cos 60 = 0.5000, tan 45 = 1.0000, sin(30 radians) = -0.9880, arcsin(0.5) = 30. degrees or 0.52359878 radians. The flags and the format share the part after ;: R%.4f is "radians, 4 decimals".

Expect: A, B, C, D, B (80 is a B: >= includes the limit).

Try: change Dora's score to 59.

Expect: F.

Because a true condition is 1 and a false one 0, a column of conditions can simply be summed to count the rows where it holds.

Expect: Low? = 0, 1, 1, 0; Low items = 2.

Expect: the solved table's values: for 2.71828 that is 2.72, 2, 3, 271.8%, 2.718e+00, 2.72, 2.72, 2.72e0.

Small numbers switch to Calc's scientific notation (4.56e-4) in the Calc formats, and %d cuts -7.5 to -7 while %.0f rounds it to -8.

Try: in the practice copy change the flag of column 7 from n3 to n5 and recalculate.

Expect: column 7 shows 2.7183, 1234.5, 4.56e-4, -7.5.

p20 asks Calc for 20 significant digits instead of 12 while computing. The display still shows at most 8 significant digits, so on its own it changes nothing you can see:

Expect: 0.66666667 and 1.4142136 in both columns. (org.nvim computes with Lua floating-point numbers, so it cannot really go beyond about 16 digits; Emacs' Calc can.)

Expect: 0.33333333, 1:3, 1:2; 0.75, 3:4, 11:12; 0.75, 3:4, 11:12 (6/8 is reduced). 1:3 is Calc's way of writing the fraction 1/3.

By default:

• an empty field used on its own counts as 0;
• an empty field inside a range is dropped (so vmean and vcount ignore it).

The E flag keeps empty fields: alone or in a range they become nan (Calc's "not a number"), which spreads to the result. EN keeps them and turns them into 0. N reads every field as a number, text included (abc is 0).

Expect: row 1 is plain: 10 everywhere, means 5, count 2.

• Row 2 (b empty): 4 (empty alone is 0), nan (E keeps it), 4; vmean 4 (the empty field is dropped), 2 with EN ((4+0)/2), and vcount 2 with E.
• Row 3 (b is the word abc): 4 + abc twice (a symbol, see "Text in formulas"), 4 with N; vmean gives the formula 2 + abc / 2, with EN 2, and vcount 2.

A date can be written directly in a formula:

Expect: 3, 88, [2026-10-26 Mon], [2026-02-28 Sat] (incmonth clamps to the end of February), 20261113.

With a time in both timestamps, the difference is a fraction of a day; multiply by 24 for hours or by 1440 for minutes.

Expect: 0.1875, 4.50, 270; 0.3542, 8.50, 510.

The formats matter here: Calc keeps a time of day to 12 digits (22:00 is day ...916667), so without them the second row shows 0.354166, 8.499984 and 509.99904, and %d (which cuts off the fraction) would give 509 minutes. %.0f rounds instead.

now() is the moment of the recalculation, so this example cannot be pre-computed.

Try: press <C-c><C-c> on the #+TBLFM: line.

Expect: the current date and time as an inactive timestamp, e.g. [2026-09-28 Mon 15:06], and the whole days left until 31 December (93 on 28 September 2026).

Expect:

• x2 (no flag!): 1:15 for 1:30 (Calc read 1:30 as the fraction 1/30 and doubled it to 1/15), 0 for 0:20 (0/20) and 13:3 for 2:05:30 (the mixed fraction 2 5/30, doubled). Always use a flag with durations.
• x2;T: 03:00:00, 00:40:00, 04:11:00.
• +45 min;T: add 2700 seconds: 02:15:00, 01:05:00, 02:50:30.
• +0:45;T: no change! In the formula text, 0:45 is still the fraction 0/45 = 0; only fields are converted. Write seconds (45*60) instead.
• the last column shows the minutes: with t the result is in hours, so multiplying the duration by 60 gives minutes as "hours": 90.00, 20.00, 125.50.

The t output unit is set by the option table_duration_custom_format ("hours" by default, or "minutes", "seconds", "days"), and table_duration_hour_zero_padding (default true) decides between 03:00:00 and 3:00:00 for T and U.

Expect: Ada Lovelace, AL, LOVELACE, 12; Alan Turing, AT, TURING, 11; Grace Hopper, GH, HOPPER, 12.

Try: in the practice copy change Grace to Margaret, Hopper to Hamilton.

Expect: Margaret Hamilton, MH, HAMILTON, 17.

Expect: 7, 12, 4, small; 11, 18, 9, big; 5, 0, 5, medium (the empty b is 0 with N; in the range $1..$2 it is dropped).

Without N, '(+ $1 $2) would try to add two strings and give #ERROR.

Lisp's format gives full control over the text of a result:

Expect: 3 x 0 = 3.60 and 10 x 0 = 5.00. With N every field is a number, so the item name Pen became 0. Mixing text and numbers needs the conversion done by hand, without N:

Expect: 3 x Pen = 3.60 and 10 x Pad = 5.00.

With L the text of the field is inserted as Lisp source. That is the only way to write a field that holds a Lisp form, and makes numbers numbers without N:

Expect: 9 and 3; 49 and 42.

Expect: one-two-three, eerht owt eno, 3, 1; red-green, neerg der, 2, 0. Note: gsub returns two values; only the first one is used, except where select(2, ...) picks the count.

Without any ! row, $Name refers to the column whose header, above the first horizontal line, is Name. Emacs does not support this; use a ! row if the file must also work there.

Expect: 7.00, 5.00, 9.90.

Try: in the practice copy swap the Qty and Price columns (cursor in Qty, <M-l>) and recalculate.

Expect: the same totals: the formula follows the names, not positions.

If the first column of a table holds only marks (!, $, ^, _, #, *, / or nothing), it is a marks column. A row marked ! names the columns under it:

Expect: 80, 480, 300, Total 860. The ! row sits above the first horizontal line, in the header, so it is not part of the range @I..@II (a word like total inside a summed range would turn the sum into the formula total + 860). The ! row itself is never changed, and with a marks column the column formulas only touch rows marked # or * (see "Automatic recalculation" below).

On the left side, Emacs accepts only column numbers ($5=, @>$5=); $total= and @>$total= are org.nvim extensions. Formulas run in the order of their left sides compared as text, and a column name counts as its number there: $total= runs where $5= would, so named column formulas run left to right, like numbered ones.

A row marked $ defines parameters as name=value fields; formulas use them as $name.

Expect: 51.04, 153.12, 382.80.

Try: in the practice copy change rate=1.1 to rate=1 and recalculate.

Expect: 46.40, 139.20, 348.00.

#+CONSTANTS: lines anywhere in the file define names for every table of the file. This file starts with

#+CONSTANTS: vat=0.20 hourly=45 g=9.81

(and table_formula_constants in your setup can define global ones). A $ row or ! name of the table wins over a constant of the same name.

Expect: Pay 360 and 562.5; with VAT 432.00 and 675.00. (A whole float like 625.0 would show as 625., Calc's way of saying "float".)

Try: change hourly=45 to hourly=50 on the #+CONSTANTS: line at the top of the file, then come back and press <C-c><C-c> on the practice #+TBLFM:.

Expect: 400, 625.; 480.00, 750.00. (Put 45 back afterwards.)

Falling objects, with g:

Expect: 9.81, 19.62, 29.43; 4.9, 19.6, 44.1.

$PROP_Name is the value of the property Name of the entry the table is in (inherited from parents too).

A ^ row names the fields of the row above it; a _ row names the fields of the row below it. Use them for single important fields such as a total. Formulas see the value a named field had before the recalculation, so a formula that uses a freshly computed named field needs a second pass (16<C-c>* iterates).

Read it as: total is the Spent field of the Total row (the ^ row names the field above it), each is the Spent field of the Each row (the _ row names the field below it). $total…= and $each…= are named field formulas. The Share column says how much each person paid above (+) or below (-) an even split. The Total and Each rows have no mark, so the column formula for Share skips them.

Try: press 16<C-c>* in the practice copy.

Expect: Total 400, Each 133.33, Share -13.33, -53.33, 66.67.

The message says "Convergence after 4 iterations".

Try: press u until the copy is empty again, then <prefix>Tf three times, watching the table after each press.

Expect:

1. Total 400, but Each and all shares #ERROR: formulas read a named field as it was before the pass, and total and each were empty.
2. Each 133.33; the shares are still #ERROR (each was empty before this pass).
3. The shares -13.33, -53.33, 66.67. A fourth pass changes nothing.

A summary table that reads from q3-sales:

Expect: 314, 133, 120 (September), 34.89, 3.

Try: change South's September to 58 in q3-sales. Recalculate q3-sales first (<C-c><C-c> on its #+TBLFM:), then the practice copy.

Expect: 334, 133, 140, 37.11, 3. If you recalculate only the summary, it still reads the old totals of q3-sales: remote() reads the other table as it is, it does not recalculate it. :Org table_recalc_buffer recalculates every table of the buffer, repeating until nothing changes, which settles chains like this one.

• Diff: positive means you spent more than planned.
• % of spent divides by the Actual total in the last row, so it needs a second pass (16<C-c>*).
• Status is a range formula for rows 2-7 (@2$6..@7$6), so the Total row gets no status.

Expect: Diff 0, 62, -25, 24, -50, 0; % 45.1, 19.2, 3.6, 7.7, 5.6, 18.8; Status ok, over, ok, over, ok, ok; Total 2650, 2661, 11, 100.0.

Try: change Fun's Actual to 260 and press 16<C-c>*.

Expect: Fun Diff 60, Status over; totals 2650, 2771, 121; the percentages all shift: 43.3, 18.5, 3.4, 7.4, 9.4, 18.0.

Uses the constant vat=0.20 from the top of the file and a $ row for the discount.

Rows: @1 header, @2 the ! row, @3-@5 items, @6 Subtotal, @7 Discount, @8 VAT, @9 Total due, @10 the $ row. The items are between lines II and III. The field formulas run in row order, so each one can use the rows above it in the same pass. The left sides use column numbers ($5, @6$5): names work on the right side, and Emacs only allows numbers on the left.

Expect: Net 1140.00, 180.00, 149.70; Subtotal 1469.70, Discount -73.48, VAT 279.24, Total due 1675.46. (5% of 1469.70 is 73.485; floating point makes it a hair less, so it rounds down. Each row uses the already rounded value of the rows above, as a printed invoice would.)

Try: change disc=0.05 to disc=0.10 and recalculate.

Expect: Discount -146.97, VAT 264.55, Total due 1587.28.

Weights: homework 30%, midterm 30%, final 40%. The average row uses vmean, the grade a nested if.

Expect: Weighted 92.0, 71.7, 53.7, 88.4, 86.3; Grade A, C, F, B, B; Average 82.0, 76.2, 77.4, 78.4; Best 98, 91, 95, 92. (the maximum of the floats 92.0 … is the float 92, written 92. because that formula has no format).

Try: Cid retakes the final and gets 78. Recalculate.

Expect: Cid 65.7 and D; the Final average becomes 83.4, the Weighted average 80.8.

• Worked = Out - In - Break, written as H:MM by the U flag.
• 8*3600 is 8 hours in seconds (with a duration flag plain numbers are seconds). Overtime can be negative.

Expect: Worked 08:00, 08:15, 07:00, 09:00, 06:00; Overtime 00:00, 00:15, -01:00, 01:00, -02:00; Week 03:00 (breaks), 38:15 (worked), -01:45 (overtime).

Try: on Friday change Out to 17:00.

Expect: Friday 09:00, overtime 01:00; Week 41:15 worked, 01:15 overtime.

@-1 alone is "the row above, this column": the previous balance. Empty In/Out fields count as 0.

Expect: 1500.00, 550.00, 465.70, 3265.70, 3204.50, 3230.00.

Try: add a row at the end: put the cursor on the Refund row, press <M-CR> (a new empty row appears below), press i and fill in <2026-10-19 Mon> <Tab> Phone <Tab> <Tab> 39.99, then <Esc> and <prefix>Tf.

Expect: the new balance 3190.01. The range @3$5..@>$5 grows with the table because it ends at @>.

The monthly payment of a loan of P at a yearly rate r over n months is P * i / (1 - (1+i)^-n) with i = r/12.

Watch the parentheses: $p*($r/1200)/(...) is fine because the division comes last; $p*$r/1200/(...) would divide by the product of 1200 and (...), which happens to be the same, but $p/(...)*... would not be.

Expect: Payment 304.22, 1013.37, 255.76; Total 10951.92, 364813.20, 15345.60; Interest 951.92, 164813.20, 345.60.

Try: change the second loan to 25 years.

Expect: Payment 1111.66, Total 333498.00, Interest 133498.00.

Compound interest year by year, each row growing the row above:

Expect: 1050.00, 1102.50, 1157.62, 1215.50, 1276.28. Each row multiplies the rounded text of the row above (the field holds 1102.50, not more digits), and 1157.625 is rounded to even by %.2f.

Try: change the rate to 7.

Expect: 1070.00, 1144.90, 1225.04, 1310.79, 1402.55.

The Action column uses a nested if with words; the Order? column is the bare condition $st < $min, which is 1 or 0, so its sum counts the items to order. (Avoid a hyphen in words: ORDER-NOW would be read as the subtraction ORDER - NOW.)

Expect: Value 33.60, 7.50, 1.80, 40.80, 0.00, Total 83.70; Action ok, reorder, reorder, ok, urgent; Order? 0, 1, 1, 0, 1, and 3 items to order.

Try: a delivery brings 300 nuts: set Nuts' Stock to 450; then, with the cursor still on the Nuts row, press <C-c><C-c> (the row is marked #).

Expect: only that row is recalculated: Value 22.50, Action ok, Order? 0. The totals still show 83.70 and 3 until <prefix>Tf makes them 98.70 and 2.

Each number is the sum of the two above it.

The ratio range starts at row 4, like the Fibonacci range. Starting it at row 3 would break it: @3$3.. sorts before @4$2.., so the ratios would be computed before the numbers exist (all 0.000000 on the first pass).

Expect: 1 1 2 3 5 8 13 21 34 55 89 144; the ratio approaches the golden ratio: the last one is 1.618056.

Try: change the two starting values to 2 and 1 (the Lucas numbers).

Expect: 2 1 3 4 7 11 18 29 47 76 123 199.

A project plan: each task starts the day after the previous one ends. The duration is in days; dates come out as inactive timestamps.

The column formula for End runs first, when the later Starts are still empty, so this sheet needs iterating: press 16<C-c>* (or <prefix>Tf five times).

Expect: Ends [2026-10-09 Fri], [2026-10-17 Sat], [2026-11-01 Sun], [2026-11-07 Sat], [2026-11-10 Tue]; Starts [2026-10-10 Sat], [2026-10-18 Sun], [2026-11-02 Mon], [2026-11-08 Sun]; Total 37 days, and the last field 37 (days from the first start to the last end).

Try: change Build to 20 days and 16<C-c>*.

Expect: Build ends [2026-11-06 Fri], Launch prep ends [2026-11-15 Sun], Total 42.

A weekly meeting series, simpler: one date plus 7 days, row after row.

Expect: [2026-10-13 Tue], [2026-10-20 Tue], [2026-10-27 Tue], [2026-11-03 Tue].

A field shows #ERROR when its formula cannot be evaluated. The other fields are still computed. Common causes, each shown in a row below:

Expect: #ERROR in every row. The fixes: close the parenthesis; use a column that exists; use @> instead of guessing the last row; define the name (! row, $ row, #+CONSTANTS:); add the N flag ('(+ $2 1);N gives 4); call a function that exists ('(math.floor($2))).

Try: in the practice copy change $2*(2 to $2*(2), $9 to $2, @20$2 to @>$2, $nope to $vat (a constant of this file), add ;N after '(+ $2 1) and change math.nothing to math.floor. Recalculate.

Expect: 6, 4, 3, 0.4 (0.20*2), 4, 3.

These give a result, just not the one you meant. All were explained in earlier sections:

Two formulas that read each other never settle. One pass computes each from the old value of the other; iterating repeats that until the limit.

Try: press <C-c><C-c> on the #+TBLFM: line, then again, then 16<C-c>*.

Expect: first 1 and 2 ($3 was empty, i.e. 0, when $2 was computed); then 3 and 4; then after ten more passes the warning "No convergence after 10 iterations" and 23 / 24. There is no right answer: rewrite one formula so it does not depend on the other.

A field formula that points beyond the last column is refused (unless table_formula_create_columns is set):

Try: <C-c><C-c> on the #+TBLFM: line.

Expect: the error "Table formula error: Missing columns in the table. Aborting", while the column formula did its job: 10 and 20. Remove ::@2$3=99 (or add a third column: cursor in column b, <prefix>Ti) and the error goes away.

1. <C-c>? on the field: is the reference the one you think, and which formula is active there? ("line @3, col $4, ref @3$4 or D3, formula: …")
2. <C-c>}: show all row and column numbers on the table.
3. <prefix>': open the formula editor; moving over a formula highlights the fields it uses.
4. <C-c>{: turn on the debugger and recalculate; the *Substitution History* window shows the formula after the references were replaced, so you see exactly what Calc was asked to compute.
5. Recalculate twice. If the second pass changes something, you have an ordering problem; 16<C-c>* solves it (or a circular reference, which it cannot).

Try: put the cursor on the Result field of the row "unknown name" in the #ERROR practice copy (before fixing it), press <C-c>{, then <C-c><C-c> on its #+TBLFM: line and step with y through the questions until you reach the $nope*2 formula.

Expect: for $nope*2 the window shows an Error: line instead of Result:. Press <C-c>{ to switch the debugger off when done.

<C-c>"a on a column of numbers adds a new column after it with a bar per row, drawn by a Lisp formula orgtbl-ascii-draw that becomes part of the #+TBLFM: line, so the bars follow the numbers when you recalculate. The bars go from the column's minimum to its maximum, 12 characters wide (a count sets another width, 4<C-c>"a asks).