sylk and xlsb_short_records

README.md

@@ -1,2 +1,10 @@
-# notes
-Various file format notes
+# SheetJS File Format Notes
+
+Various spreadsheet file format notes.
+
+- [Symbolic Link (SLK/SYLK)](/sylk/README.md)
+- [XLSB Short Records](/xlsb_short_records/README.md)
+
+Project sponsored by [SheetJS](https://sheetjs.com)
+
+[![Analytics](https://ga-beacon.appspot.com/UA-36810333-1/SheetJS/notes?pixel)](https://github.com/SheetJS/notes)

_config.yml

@@ -0,0 +1 @@
+title: SheetJS File Format Notes

sylk/README.md

@@ -0,0 +1,83 @@
+# Symbolic Link format
+
+Files start with `ID` (`0x49 0x44`).  Files are interpreted as plaintext in the
+system ANSI codepage.
+
+
+## Basics
+
+The file consists of a series of plaintext records.  Records are separated by
+newline characters (both `\r\n` and `\n` newlines are accepted by newer versions
+of Excel, but generated files should prefer CRLF).
+
+### Fields
+
+A record consists of a record type and a series of fields.  Each part of the
+record is separated by a single `;` character.
+
+The literal semicolon is encoded as two consecutive semicolons `;;`.  Example:
+
+```
+C;Y1;X1;K"abc;;def"
+```
+
+### Encoding
+
+In addition to the escaped semicolon, Excel understand two types of Encodings:
+
+#### Raw Byte Trigrams
+
+Trigrams matching the pattern `\x1B[\x20-\x2F][\x30-\x3F]` are decoded into a
+single byte whose high bits are taken from the second character and whose low
+bits are taken from the third character.
+
+For example. `"\x1B :" == "\x1B\x20\x3A` encodes the byte `"\x0A"` (newline)
+
+`"\x1B#;` encodes a literal semicolon.
+
+#### Special Escapes
+
+Excel also understands a set of special escapes that start with `\x1BN`.  For
+clarity, the `\x1BN` part is not included in the table:
+
+| sequence | text |
+|:---------|:-----|
+| `AA`     | `À`  |
+
+
+## Record Types
+
+| Record Type | Description          |
+|:------------|:---------------------|
+| `ID`        | Header               |
+| `E`         | EOF                  |
+| `B`         | Worksheet Dimensions |
+| `O`         | Options              |
+| `P`         | Number Format        |
+| `F`         | Formatting           |
+| `C`         | Cell                 |
+
+
+## EOF Record (E)
+
+There are no fields.
+
+
+## Cell Record (C)
+
+
+### Comments
+
+The `A` field of the `C` record can specify plaintext comments. They are encoded
+using the same text encoding in `K` fields.
+
+### Shared Formulae
+
+The `S` field of the `C` record signals that a cell is using a shared formula.
+The `R` and `C` fields are the 1-indexed row and column indices of the cell with
+the formula.  The formula should be extracted from the original location and
+shifted to the current cell (relative references adjusted by the offset).
+
+
+
+[![Analytics](https://ga-beacon.appspot.com/UA-36810333-1/SheetJS/notes?pixel)](https://github.com/SheetJS/notes)

sylk/comment.slk

@@ -0,0 +1,10 @@
+ID;PWXL;N;E
+P;PGeneral
+F;P0;DG0G10;M320
+B;Y3;X1;D0 0 9 0
+C;Y1;X1;AArthas: :I would gladly bear any curse to save my homeland.
+C;Y2;X2;AMuradin: :Leave it be, Arthas.  Forget this business and lead your men home.
+C;Y1;X1;K1
+C;Y1;X2;K2
+C;Y2;X1;K3
+E

sylk/shared_formula.slk

@@ -0,0 +1,8 @@
+ID;PWXL;N;E
+P;PGeneral
+F;P0;DG0G10;M320
+B;Y3;X1;D0 0 9 0
+C;Y1;X1;K1
+C;Y2;K2;ER[-1]C+1
+C;Y3;K3;S;R2;C1
+E

xlsb_short_records/README.md

@@ -0,0 +1,90 @@
+# XLSB Short Records
+
+There are 7 undocumented XLSB records (record types 12-18) that Excel supports.
+They appear to specify cells using a "Short" cell structure
+
+## Cell Structures
+
+XLSB Cell structures are 8 bytes with the following layout:
+
+```
+column index (4 bytes)
+style index (3 bytes)
+flags (1 byte)
+```
+
+A "Short" structure is 4 bytes and omits the column:
+
+```
+style index (3 bytes)
+flags (1 byte)
+```
+
+The actual column index is understood to be the column after the previous cell.
+For example, if D3 was the last cell, a record using the Short structure is
+defining cell E3.
+
+## Cell Records
+
+The various cell records (BrtCellBlank, BrtCellBool, etc) consist of a Cell
+structure followed by the cell data.  The various formula records (BrtFmlaBool,
+BrtFmlaError, etc) append the formula structure to the base cell record.
+
+The "Short" cell records follow similar patterns but omit the 4-byte column
+field from the cell structure.
+
+For example, record type 18 "BrtShortIsst" is the short form of BrtCellIsst.
+
+BrtCellIsst has the following layout:
+
+```
+column index (4 bytes)
+style index (3 bytes)
+flags (1 byte)
+shared string table index (4 bytes)
+```
+
+BrtShortIsst omits the column index:
+
+```
+style index (3 bytes)
+flags (1 byte)
+shared string table index (4 bytes)
+```
+
+## Records
+
+| Record | Name          | Long Cell Record |
+|-------:|:--------------|:-----------------|
+|   `12` | BrtShortBlank | BrtCellBlank     |
+|   `13` | BrtShortRk    | BrtCellRk        |
+|   `14` | BrtShortError | BrtFmlaError     |
+|   `15` | BrtShortBool  | BrtCellBool      |
+|   `16` | BrtShortReal  | BrtCellReal      |
+|   `17` | BrtShortSt    | BrtCellSt        |
+|   `18` | BrtShortIsst  | BrtCellIsst      |
+
+Record 13 is informally referred to as "BrtShortRk".  It is the short form of
+BrtCellRk.  BrtCellRk is a 12 byte structure:
+
+```
+column index (4 bytes)
+style index (3 bytes)
+flags (1 byte)
+value stored as RkNumber (4 bytes)
+```
+
+The short form BrtShortRk is therefore an 8 byte structure:
+
+```
+style index (3 bytes)
+flags (1 byte)
+value stored as RkNumber (4 bytes)
+```
+
+## Test Files
+
+- [`brt_str.xlsb`](./brt_str.xlsb) includes types 12,13,14,15,16,17
+- [`brt_sst.xlsb`](./brt_sst.xlsb) includes types 12,13,14,15,16,18
+
+[![Analytics](https://ga-beacon.appspot.com/UA-36810333-1/SheetJS/notes?pixel)](https://github.com/SheetJS/notes)