numbers-parser
numbers-parser is a Python module for parsing Apple Numbers .numbers files. It supports Numbers files generated by Numbers version 10.3, and all 11.x up to 11.2 (current as of November 2021).
It supports and is tested against Python versions from 3.6 onwards. It is not compatible with earlier versions of Python.
Currently supported features of Numbers files are:
- Multiple sheets per document
- Multiple tables per sheet
- Text, numeric, date, currency, duration, percentage cell types
Formulas have very limited support and rely wholly on Numbers saving values in cells as part of the saved document, which is not always guaranteed. When a formula value is not present, the value *FORMULA* is returned. Any formula that results in a Numbers error returns a value *ERROR*.
Installation
python3 -m pip install numbers-parser
Usage
Reading documents:
from numbers_parser import Document
doc = Document("my-spreasdsheet.numbers")
sheets = doc.sheets()
tables = sheets[0].tables()
rows = tables[0].rows()
Referring to sheets and tables
Both sheets and names can be accessed from lists of these objects using an integer index (list syntax) and using the name of the sheet/table (dict syntax):
# list access method
sheet_1 = doc.sheets()[0]
print("Opened sheet", sheet_1.name)
# dict access method
table_1 = sheets["Table 1"]
print("Opened table", table_1.name)
Accessing data
Table objects have a rows method which contains a nested list with an entry for each row of the table. Each row is itself a list of the column values. Empty cells in Numbers are returned as None values.
data = sheets["Table 1"].rows()
print("Cell A1 contains", data[0][0])
print("Cell C2 contains", data[2][1])
Cell references
In addition to extracting all data at once, individual cells can be referred to as methods
doc = Document("my-spreasdsheet.numbers")
sheets = doc.sheets()
tables = sheets["Sheet 1"].tables()
table = tables["Table 1"]
# row, column syntax
print("Cell A1 contains", table.cell(0, 0))
# Excel/Numbers-style cell references
print("Cell C2 contains", table.cell("C2"))
Merged cells
When extracting data using data() merged cells are ignored since only text values are returned. The cell() method of Table objects returns a Cell type object which is typed by the type of cell in the Numbers table. MergeCell objects indicates cells removed in a merge.
doc = Document("my-spreasdsheet.numbers")
sheets = doc.sheets()
tables = sheets["Sheet 1"].tables()
table = tables["Table 1"]
cell = table.cell("A1")
print(cell.merge_range)
print(f"Cell A1 merge size is {cell.size[0]},{cell.size[1]})
Row and column iterators
Tables have iterators for row-wise and column-wise iteration with each iterator returning a list of the cells in that row or column
for row in table.iter_rows(min_row=2, max_row=7, values_only=True):
sum += row
for col in table.iter_cole(min_row=2, max_row=7):
sum += col.value
Pandas
Since the return value of data() is a list of lists, you should be able to pass it straight to pandas like this
import pandas as pd
doc = Document("simple.numbers")
sheets = doc.sheets()
tables = sheets[0].tables()
data = tables[0].rows(values_only=True)
df = pd.DataFrame(data, columns=["A", "B", "C"])
Numbers File Formats
Numbers uses a proprietary, compressed binary format to store its tables. This format is comprised of a zip file containing images, as well as Snappy-compressed Protobuf .iwa files containing metadata, text, and all other definitions used in the spreadsheet.
Protobuf updates
As numbers-parser includes private Protobuf definitions extracted from a copy of Numbers, new versions of Numbers will inevitably create .numbers files that cannot be read by numbers-parser. As new versions of Numbers are released, the following steps must be undertaken:
- Run proto-dump on the new copy of Numbers to dump new Proto files.
- proto-dump assumes version 2.5.0 of Google Protobuf which may need changes to build on more modern OSes. The version linked here is maintained by the author and tested on recent macOS for both arm64 and x86_64 architectures.
- Any
.characters in the Protobuf definitions must be changed to_characters manually, or via therename_proto_files.pyscript in theprotosdirectory of this repo.
- Connect to a running copy of
Numberswithlldb(or any other debugger) and manually copy and reformat the results ofpo [TSPRegistry sharedRegistry]intomapping.py.- Versions of macOS >= 10.11 may protect Numbers from being attached to by a debugger - to attach, temporarily disable System IntegrityProtection to get this data.
- The
generate_mapping.pyscript inprotosshould help turn the output from this step into a recreation ofmapping.py
Running make bootstrap will perform all of these steps and generate the Python protos files as well as mapping.py. The makefile assumes that proto-dump is in a repo parallel to this one, but the make variable PROTO_DUMP can be overridden to pass the path to a working version of proto-dump.
Credits
numbers-parser was built by Jon Connell but derived enormously from prior work by Peter Sobot. Both modules are derived from previous work by Sean Patrick O'Brien.
Decoding the data structures inside Numbers files was helped greatly by previous work by Steven Lott.
Formula tests were adapted from JavaScript tests used in fast-formula-parser.
License
All code in this repository is licensed under the MIT License
1 Jan 18, 2022
3 Jan 24, 2022
1 Oct 18, 2021
7 Nov 11, 2021
6 Dec 08, 2022
86 Dec 19, 2022
18 Dec 23, 2022
4 Jun 24, 2022
2 Jan 10, 2022
18 Nov 15, 2021
11 Jul 30, 2021
120 Dec 05, 2022
4 Sep 12, 2021
23 Jan 15, 2022
1 Dec 07, 2021
29 Dec 05, 2022
4 Dec 01, 2022
13 Nov 25, 2022
4 Aug 22, 2022
1 Oct 04, 2022