My thinking about these attributes is that position refers to a single point on the shape, and [xanchor, yanchor] (plus padding) refers to a single point on the text element, and you position the text so that those two points are in the same place.

So for lines, the only position values that make sense are start|middle|end. We can set smart defaults for xanchor based on those ({start: 'left', middle: 'center', end: 'right'} so that by default for position: start|end the start or end of the text aligns with the start or end of the line, and for position: middle the text is centered on the line. Then yanchor: bottom puts the text on top of the line (ie the bottom of the text is on the line), yanchor: top puts the text under the line, and yanchor: 'middle' puts the text on top of the line. You can get other cool effects then like text preceding the line (position: start, xanchor: right, yanchor: middle) or following the line (position: end, xanchor: left, yanchor: middle)

For other shapes the first nine positions you have are good, and in those cases we can set smart defaults for both xanchor and yanchor to match, ie [xanchorDefault, yanchorDefault] = position.split(' ') so (for rects anyway) by default the text is drawn just inside whatever corner or side you specify with position. Default for all these shapes should be middle center though.