Instead of re-encoding the full sheet to change the namespaces
in the encoded bytes, read the sheet again and do the byte
replacements there.
In this case, file access ends up being more performant than
marshaling the sheet back to XML.
In the SharedStrings test, ensure the strings are actually read.
Fix#439
* Huge refactorig for consistent col/row numbering
Started from simply changing ToALphaString()/TitleToNumber() logic and related fixes.
But have to go deeper, do fixes, after do related fixes and again and again.
Major improvements:
1. Tests made stronger again (But still be weak).
2. "Empty" returns for incorrect input replaces with panic.
3. Check for correct col/row/cell naming & addressing by default.
4. Removed huge amount of duplicated code.
5. Removed ToALphaString(), TitleToNumber() and it helpers functions at all,
and replaced with SplitCellName(), JoinCellName(), ColumnNameToNumber(), ColumnNumberToName(), CellNameToCoordinates(), CoordinatesToCellName().
6. Minor fixes for internal variable naming for code readability (ex. col, row for input params, colIdx, rowIdx for slice indexes etc).
* Formatting fixes
* Implement consistent row addressing by Excel row number starting with 1
1. Added second versions for all row manipulation methods with zero-based row addressing.
2. Fixed methods documentation to explicitly describe which row addressing used in method.
3. Added WARNING to README.md.
4. Cosmetic change: All row test moved to file `rows_test.go`.
* TravisCI: go1.12 added to tests matrix
* BACKWARD INCOMPARTIBLE: Use only Excel numbering logic from 1 row
* README updated
* TestTitleToNumber: more test cases
* TitleToNumber: drop use of math.Pow()
Compute using pure integers
* TitleToNumber: simplify
Remove unecessary casts to int
DuplicateRowTo() is similar to DuplicateRow() but copies specified row not just after specified source row
but to any other specified position below or above source row.
Also I made minor modifications of tests: using filepath.Join() instead of direct unix-way paths strings
to avoid possible tests fails on other OS.
Primary motivation: Avoid statefull tests with not ignorable git file tree changes.
Multiple tests reads and overwrites signle file for won needs.
Multiple tests reads and overwrites file under version control.
Secondary motivation: Minimal tests logic aligment, separate error expectation
and not error expectation tests. Introduce sub-test over test data sets and so far.
This commit is not ideal but necessary (IMHO)
Use github.com/mohae/deepcopy to deep copy worksheets instead of the
internal deepcopy function that was using encoding/gob serialization and
deserialization.
Rationale:
1/ using `encoding/gob` is much slower than
[`mohae/deepcopy`](https://github.com/mohae/deepcopy/)
2/ When building an application this implementation of `deepcopy` drags
the `encoding/gob` package into the binary. And this package is much
bigger than `mohae/deepcopy` (which only depends on `time` and `reflect`).
```
$ LC_ALL=C stat -f "%6z %N" $(go env GOPATH)/pkg/$(go env GOOS)_$(go env GOARCH)/github.com/mohae/deepcopy.a $(go env GOROOT)/pkg/$(go env GOOS)_$(go env GOARCH)/encoding/gob.a
10508 .../pkg/darwin_amd64/github.com/mohae/deepcopy.a
541818 .../pkg/darwin_amd64/encoding/gob.a
```
Harris
xuri