[open] handle input files that cannot seek backward

visidata/_open.py

@@ -93,8 +93,6 @@ def openPath(vd, p, filetype=None, create=False):
     filetype = filetype.lower()
 
     if not p.exists():
-        if not create:
-            return None
         newfunc = getattr(vd, 'new_' + filetype, vd.getGlobals().get('new_' + filetype))
         if not newfunc:
             vd.warning('%s does not exist, creating new sheet' % p)
@@ -106,6 +104,9 @@ def openPath(vd, p, filetype=None, create=False):
     openfuncname = 'open_' + filetype
     openfunc = getattr(vd, openfuncname, vd.getGlobals().get(openfuncname))
     if not openfunc:
+        if not p.seekable():
+            # read the file as text, into a RepeatFile that can be opened multiple times
+            p = Path(p.given, fp=p.open(mode='rb'))
         opts = vd.guessFiletype(p)
         if opts and 'filetype' in opts:
             filetype = opts['filetype']
@@ -130,7 +131,6 @@ def openPath(vd, p, filetype=None, create=False):
 
     return openfunc(p)
 
-
 @VisiData.api
 def openSource(vd, p, filetype=None, create=False, **kwargs):
     '''Return unloaded sheet object for *p* opened as the given *filetype* and with *kwargs* as option overrides. *p* can be a Path or a string (filename, url, or "-" for stdin).
@@ -164,6 +164,9 @@ def openSource(vd, p, filetype=None, create=False, **kwargs):
 def open_txt(vd, p):
     'Create sheet from `.txt` file at Path `p`, checking whether it is TSV.'
     if p.exists(): #1611
+        if not p.seekable():
+            # read the file as text, into a RepeatFile that can be opened multiple times
+            p = Path(p.given, fp=p.open(mode='rb'))
         with p.open(encoding=vd.options.encoding) as fp:
             delimiter = vd.options.delimiter
             try:

visidata/path.py

@@ -197,6 +197,8 @@ def open(self, mode='rt', encoding=None, encoding_errors=None, newline=None):
         # rfile makes a single-access fp reusable
 
         if self.rfile:
+            if 'b' in mode:
+                raise ValueError('a RepeatFile holds text and cannot be reopened in binary mode')
             return self.rfile.reopen()
 
         if self.fp:
@@ -291,6 +293,12 @@ def read_bytes(self):
         with self.open(mode='rb') as fp:
             return fp.read()
 
+    # Some unusual paths can be read from, but do not allow seeking.
+    # For example, a pipe created by process substitution in bash, or a symlink to a fifo.
+    def seekable(self):
+        with self.open() as fp:
+            return fp.seekable()
+
     def is_url(self):
         'Return True if the given path appears to be a URL.'
         return '://' in self.given
@@ -376,9 +384,19 @@ def tell(self):
         '''Tells the current position as an opaque line marker.'''
         return self.iter.nextIndex
 
-    def seek(self, n):
+    def seek(self, offset, whence=io.SEEK_SET):
         '''Seek to an already seen opaque line position marker only.'''
-        self.iter.nextIndex = n
+        if whence != io.SEEK_SET and offset != 0:
+            if whence == io.SEEK_CUR:
+                raise io.UnsupportedOperation("can't do nonzero cur-relative seeks")
+            elif whence == io.SEEK_END:
+                raise io.UnsupportedOperation("can't do nonzero end-relative seeks")
+            else:
+                raise ValueError('invalid whence (%s, should be %s, %s or %s)' % (whence, io.SEEK_SET, io.SEEK_CUR, io.SEEK_END))
+        self.iter.nextIndex = offset
+
+    def seekable(self):
+        return True
 
     def readline(self, size=-1):
         if size != -1: