Export library code in pkg/ (#1391)

* Export library code in `pkg/`

* new doc page

pkg/platform/README.md

@@ -0,0 +1 @@
+This is Miller's platform-dependent code -- as of April 2021, all Windows vs not-Windows.

pkg/platform/diff_notwindows.go

@@ -0,0 +1,14 @@
+// ================================================================
+// Handling diff or fc for regression-test.
+// ================================================================
+
+//go:build !windows
+// +build !windows
+
+package platform
+
+// GetDiffRunArray gets the command for diffing actual/expected output in a
+// regression test.
+func GetDiffRunArray(filename1, filename2 string) []string {
+	return []string{"diff", "-u", filename1, filename2}
+}

pkg/platform/diff_windows.go

@@ -0,0 +1,14 @@
+// ================================================================
+// Handling diff or fc for regression-test.
+// ================================================================
+
+//go:build windows
+// +build windows
+
+package platform
+
+// GetDiffRunArray gets the command for diffing actual/expected output in a
+// regression test.
+func GetDiffRunArray(filename1, filename2 string) []string {
+	return []string{"fc", filename1, filename2}
+}

pkg/platform/doc.go

@@ -0,0 +1,3 @@
+// Package platform is Miller's platform-dependent code -- as of April 2021,
+// this is divided into Windows and not-Windows.
+package platform

pkg/platform/getargs_notwindows.go

@@ -0,0 +1,17 @@
+// ================================================================
+// Handling single quotes and double quotes is different on Windows unless
+// particular care is taken, which is what this file does.
+// ================================================================
+
+//go:build !windows
+// +build !windows
+
+package platform
+
+import (
+	"os"
+)
+
+func GetArgs() []string {
+	return os.Args
+}

pkg/platform/getargs_windows.go

@@ -0,0 +1,154 @@
+// ================================================================
+// Handling single quotes and double quotes is different on Windows unless
+// particular care is taken, which is what this file does.
+// ================================================================
+
+//go:build windows
+// +build windows
+
+package platform
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+
+	shellquote "github.com/kballard/go-shellquote"
+	"golang.org/x/sys/windows"
+)
+
+// GetArgs returns a copy of os.Args, as-is except for any arg wrapped in single quotes.
+// This is for compatibility Linux/Unix/MacOS.
+//
+// For example:
+//
+//   mlr --icsv --ojson put '$filename = $basename . ".ext"' ..\data\foo.csv
+//
+// The Windows way to say that is
+//
+//   mlr --icsv --ojson put "$filename = $basename . """.ext"""" ..\data\foo.csv
+//
+// This function makes it possible to say the former, or the latter.
+
+func GetArgs() []string {
+	// If this code is running in MSYS2: MSYS2 does the right thing already and
+	// we won't try to improve on that.  This is true regardless of whether we
+	// were compiled inside MSYS2 or outside; what matters is whether we're
+	// _running_ inside MSYS2 or outside. (I.e.  this is necessarily a run-time
+	// check, not a compile-time check.)
+	msystem := os.Getenv("MSYSTEM")
+	if msystem != "" {
+		return os.Args
+	}
+
+	//printArgs(os.Args, "ORIGINAL")
+
+	regrouped, ok := regroupForSingleQuote(os.Args)
+	if !ok {
+		return os.Args
+	}
+	//printArgs(regrouped, "REGROUPED")
+
+	// Things on the command line include: args[0], which is fine as-is;
+	// various flags like -x or --xyz which are fine as-is; DSL expressions
+	// such as '$a = $b . "ccc"'.  We don't want to give back the result of
+	// shellquote.Split since that will remove the backslashes from things like
+	// ..\data\foo\dat or C:\foo\bar.baz.
+	rawCommandLine := windows.UTF16PtrToString(windows.GetCommandLine())
+	splitArgs, err := shellquote.Split(rawCommandLine)
+	if err != nil {
+		fmt.Fprintf(
+			os.Stderr,
+			"mlr: internal error: could not parse Windows raw command line: %v\n",
+			err,
+		)
+	}
+
+	retargs := make([]string, 0)
+
+	// TODO err/stetret if lens uneq
+
+	for i, oldArg := range regrouped {
+		if strings.HasPrefix(oldArg, "'") && strings.HasSuffix(oldArg, "'") {
+			retargs = append(retargs, splitArgs[i])
+		} else {
+			retargs = append(retargs, regrouped[i])
+		}
+	}
+	//printArgs(retargs, "NEW")
+
+	globbed := make([]string, 0)
+	for i, _ := range retargs {
+		// Expand things like *.csv
+		matches, err := filepath.Glob(retargs[i])
+		if matches != nil && err == nil {
+			globbed = append(globbed, matches...)
+		} else {
+			globbed = append(globbed, retargs[i])
+		}
+	}
+	//printArgs(globbed, "NEW")
+
+	return globbed
+}
+
+// ----------------------------------------------------------------
+func printArgs(args []string, description string) {
+	fmt.Printf("%s:\n", description)
+	for i, arg := range args {
+		fmt.Printf("%d %s\n", i, arg)
+	}
+	fmt.Println()
+}
+
+// ----------------------------------------------------------------
+func regroupForSingleQuote(inargs []string) ([]string, bool) {
+	outargs := make([]string, 0, len(inargs))
+	inside := false
+	var concat string
+
+	// TODO: comment
+	// TODO: UT this all
+	for _, inarg := range inargs {
+		if !inside {
+			if !strings.HasPrefix(inarg, "'") {
+				// Current arg is not single-quoted, and not inside a single-quoted region
+				outargs = append(outargs, inarg)
+
+			} else {
+
+				// Start of single-quoted region
+				if strings.HasSuffix(inarg, "'") {
+					// Start and end of single-quoted region, like '$y=$x'
+					outargs = append(outargs, inarg)
+				} else {
+					// Start but not end of single-quoted region, like '$y=
+					inside = true
+					concat = inarg
+				}
+
+			}
+
+		} else {
+			if !strings.HasSuffix(inarg, "'") {
+				// Continuation of single-quoted region
+				concat = concat + " " + inarg
+
+			} else {
+				// End of single-quoted region
+				inside = false
+				concat = concat + " " + inarg
+				outargs = append(outargs, concat)
+				concat = ""
+			}
+		}
+	}
+
+	// TODO: error not bool?
+	if inside {
+		return nil, false
+	}
+
+	return outargs, true
+}

pkg/platform/shellrun_notwindows.go

@@ -0,0 +1,12 @@
+// ================================================================
+// Wraps 'sh -c foo bar' or 'cmd /c foo bar', nominally for regression-testing.
+// ================================================================
+
+//go:build !windows
+// +build !windows
+
+package platform
+
+func GetShellRunArray(command string) []string {
+	return []string{"/bin/sh", "-c", command}
+}

pkg/platform/shellrun_windows.go

@@ -0,0 +1,51 @@
+// ================================================================
+// Wraps 'sh -c foo bar' or 'cmd /c foo bar', nominally for regression-testing.
+// ================================================================
+
+//go:build windows
+// +build windows
+
+package platform
+
+import (
+	"os"
+)
+
+// PowerShell or CMD?
+//
+// * Either PowerShell or CMD is fine for everything except the '...' in inline
+//   put/filter statements.
+//
+// * PowerShell allows fractionally more, e.g. in mlr put '$flag = $x > 10' CMD
+//   sees the ">" as a file-redirect (before our main() is ever entered) and
+//   PowerShell lets it be operator it is intended to be.
+//
+// * Neither of them allows multi-line '...' inputs.
+//
+// * Both of the previous mean that a large number of regression tests need to
+//   use mlr -f .../cases/.../0047.mlr .../cases/.../0047.input and PowerShell
+//   doesn't help us eliminate this.
+//
+// * PowerShell has about a 2-second startup time per invocation so if we
+//   use it for regression testing, we'd have to move away from one shell
+//   invocation per Miller invocation and back to batching lots of Miller
+//   statements into a file, doing diff/findstr to check status. That was the
+//   case for Miller's original bash/file regtest framework and it was hard to
+//   debug.
+//
+// In conclusion, we stick with CMD because it's faster, and PowerShell while
+// more powerful isn't *sufficiently* more powerful to justify the
+// batching-complexity to overcome its startup-latency overhead.
+
+func GetShellRunArray(command string) []string {
+	if os.Getenv("MSYSTEM") != "" {
+		// Running inside MSYS2; sufficiently Unix-like already.
+		return []string{"/bin/sh", "-c", command}
+	} else {
+		cmd := os.Getenv("COMSPEC")
+		if cmd == "" {
+			cmd = "C:\\Windows\\System32\\cmd.exe"
+		}
+		return []string{cmd, "/c", command}
+	}
+}

pkg/platform/terminal_notwindows.go

@@ -0,0 +1,7 @@
+//go:build !windows
+// +build !windows
+
+package platform
+
+func EnableAnsiEscapeSequences() {
+}

pkg/platform/terminal_windows.go

@@ -0,0 +1,15 @@
+//go:build windows
+// +build windows
+
+package platform
+
+import (
+	"syscall"
+
+	sequences "github.com/nine-lives-later/go-windows-terminal-sequences"
+)
+
+func EnableAnsiEscapeSequences() {
+	sequences.EnableVirtualTerminalProcessing(syscall.Stdout, true)
+	sequences.EnableVirtualTerminalProcessing(syscall.Stderr, true)
+}